Component automatic modeling and engineering quantity calculation method and system based on drawing recognition

By establishing the relationship between views and component entities in bridge engineering drawings, and integrating and verifying parameters, the problem of scattered and incomplete parameter identification in bridge engineering was solved. This enabled the automatic generation of three-dimensional models of bridge components and the automatic calculation of concrete quantities, improving efficiency and accuracy.

CN123069862APending Publication Date: 2026-11-10NINGBO HIGHWAY CONSTR MANAGEMENT CENT +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611172160.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-04
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing bridge engineering drawings, the parameters of the same component are scattered across multiple engineering views, making it difficult to link and integrate them. This results in incomplete parameters, numerous contradictions, and an inability to directly drive 3D modeling, leading to low efficiency and a high risk of errors.

Method used

By establishing the association between drawing views and component entities, integrating parameters from different views, and performing consistency verification based on component geometric constraints, complete component parameter data is generated, enabling automatic modeling and quantity calculation.

Benefits of technology

It improved the automation level and recognition accuracy of 3D model generation for bridge components, reduced manual data entry time, increased the efficiency of engineering quantity calculation, and reduced the error rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069862A_ABST
    Figure CN123069862A_ABST
Patent Text Reader

Abstract

The application discloses a component automatic modeling and engineering quantity calculation method and system based on drawing recognition, acquires original engineering project pictures, and performs parameter labeling on the original engineering project pictures to construct a parameter labeling data set; a target detection model is constructed, the target detection model is trained through the parameter labeling data set, and the trained model is deployed, and an engineering drawing to be recognized is input into the deployed target detection model to extract engineering labeling data; based on the engineering labeling data, a component graphic area in the engineering drawing is located, and a structure type thereof is recognized, structured parameter data is obtained through content recognition of the engineering labeling data, structured component data is obtained according to component graphic area information, the structure type and corresponding structured parameter data; the structured component data is matched with a parameter mapping rule library to generate component modeling parameter data; and automatic modeling and engineering quantity calculation and generation of a light component model are performed based on the component modeling parameter data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of engineering drawing information recognition and building information modeling automatic modeling technology, specifically involving a method and system for automatic component modeling and engineering quantity calculation based on drawing recognition. Background Technology

[0002] In bridge construction, the volume of concrete work typically needs to be calculated based on the component dimensions in the design drawings. These drawings usually contain information about the type of bridge components and their corresponding dimensions, such as the length, width, height, or diameter of components like cap beams, piers, abutments, and pile foundations. In actual project management and cost calculation, technicians need to organize the parameters of various components based on the information in the drawings and calculate the concrete volume using the appropriate formulas.

[0003] In existing technologies, the creation of bridge structural models typically relies on manual methods. Technicians need to read the component names and their dimensional parameters one by one from design drawings or scanned images, then manually input the corresponding parameters into Building Information Modeling (BIM) software to generate a 3D model, and obtain component volume data through model analysis tools. This method is not only inefficient when dealing with a large number of engineering drawings or complex structures, but is also prone to inconsistencies in parameters due to manual input errors, thus affecting the accuracy of quantity calculations.

[0004] However, due to the complexity of bridge structures, the complete structural parameters of the same component are usually not concentrated in a single drawing, but exist separately in different engineering views such as plan views, elevation views, cross-section views, and detailed views. For example, the cross-sectional dimensions of the same pier component may be represented by a cross-section view, while its height, connection relationships, and other parameters are represented by an elevation view. There are corresponding relationships between the components and geometric constraints between the parameters in different views.

[0005] Existing image recognition-based engineering drawing processing methods typically extract parameters for a single drawing area, making it difficult to determine whether annotations in different drawing views belong to the same component, and also difficult to fuse parameters scattered across multiple views. Therefore, in complex bridge engineering scenarios, even if parameter information in the drawings can be identified, problems still exist such as incomplete component parameters, contradictions between parameters, and the inability to directly use them for 3D model generation.

[0006] In recent years, with the development of image recognition technology, some technical solutions have begun to attempt to automatically identify annotation information in engineering drawings to assist in extracting parameter data. However, in existing applications, the identified drawing parameters usually still need to be manually organized and verified before being manually input into the modeling system to complete model construction. There are still many manual steps between drawing information recognition and 3D model generation, and a lack of an integrated technical solution that can automatically extract drawing parameters, perform parameter structuring, and drive 3D modeling results in relatively low efficiency in bridge engineering quantity calculation and model construction.

[0007] Therefore, it is necessary to propose a technical solution that can automatically identify component parameters from bridge engineering drawings and automatically map the identified parameter information to the component parameters in the modeling system, thereby realizing the automatic generation of three-dimensional models of bridge components and the automatic calculation of concrete quantities. Summary of the Invention

[0008] To address the problems of scattered component parameters across multiple engineering views in existing bridge engineering drawings, difficulty in associating and fusing parameters from different views, and the inability of recognition results to directly drive 3D modeling, this invention establishes a relationship between drawing views and component entities, fuses component parameters from different views, and performs parameter consistency verification based on component geometric constraints. Ultimately, it generates complete component parameter data for automatic modeling, thereby improving automation and recognition accuracy while reducing manual data entry time. The technical solution adopted by this invention is as follows:

[0009] The automatic component modeling and quantity calculation method based on drawing recognition includes the following steps:

[0010] Obtain the original engineering project drawings and their parameter annotations to construct a parameter annotation dataset;

[0011] A target detection model is constructed, and the target detection model is trained using the parameter-annotated dataset. The trained target detection model is then used to detect targets on the engineering drawings to be identified, so as to extract the engineering annotation data.

[0012] Based on engineering annotation data, the component graphic areas in the engineering drawings are located and the component structure type is identified; by performing content recognition on the engineering annotation data, structured parameter data containing parameter names and parameter values ​​is obtained; further, based on the view information, component number information and component spatial position relationship in the engineering drawings, the component association relationship between different engineering views is established, and the structured parameter data in multiple views belonging to the same entity component are merged to form structured component data containing a complete set of parameters and geometric constraint relationships;

[0013] The structured component data is matched with the parameter mapping rule base to generate component modeling parameter data;

[0014] Based on component modeling parameter data, automatic component modeling and quantity calculation are performed.

[0015] Furthermore, the target detection model is trained using a generalized intersection-union ratio (GUU) loss function. The GUU is calculated by subtracting the area of ​​the smallest closed rectangle between the predicted and ground truth boxes from the area of ​​the intersection of the predicted and ground truth boxes, then comparing the absolute value of the difference with the absolute value of the area of ​​the smallest closed rectangle, and finally subtracting the ratio from the GUU of the predicted and ground truth box areas to obtain the GUU of the predicted and ground truth box areas.

[0016] Furthermore, the engineering annotation data, including the bounding box position, annotation category, and confidence information of the parameter annotation targets, is sorted according to the confidence level to obtain a set of candidate parameter annotation targets, and finally a set of annotation detection results containing annotation category, bounding box position, and confidence information is obtained;

[0017] The annotation categories in the engineering annotation data include parameter number annotations. Based on these parameter number annotations, connected component analysis is performed on the engineering drawings. By extracting the largest connected region in the drawing, the outline of the component graphic is determined. Furthermore, according to the spatial relationship between the center point of the annotation target and the component graphic region, annotation information belonging to the same component region is categorized to identify the construction graphic region corresponding to the parameter number annotation, resulting in a set of component graphic regions and their corresponding annotation information. Existing drawing recognition technologies typically only recognize single annotation information, making it difficult to determine the correspondence between annotations and specific components. This invention, through connected component analysis and spatial relationship judgment, achieves automatic association between parameter annotations and component graphics, improving the structured nature of the parameter recognition results.

[0018] Furthermore, the annotation categories in the engineering annotation data also include dimension parameter annotations and length parameter annotations. Through the component graphic region and its annotation information set, spatial relationship analysis is performed on the engineering annotation data within the same component region. Using the parameter number annotation as a reference point, the distance between it and the dimension and length annotations is calculated respectively. The corresponding parameters are determined according to the principle of minimum distance, resulting in a parameter combination containing the parameter number, dimension parameter, and length parameter corresponding to the constructed graphic region, which is used as the component graphic region information. Compared to traditional image recognition methods that only output independent annotation information, this invention automatically establishes the logical relationship between parameters through a spatial relationship matching algorithm and completes the parameter in case of missing parameters, improving the completeness and reliability of the parameter recognition results.

[0019] Furthermore, in the spatial relationship analysis, when the detection results lack size parameters or length parameters, a reference position is constructed by analyzing the spatial arrangement relationship between existing parameters, and the annotation with the smallest distance from the reference position is selected as a supplementary parameter in the candidate annotation area.

[0020] Furthermore, based on the component numbers, bridge span numbers, station numbers, axis relationships, and spatial positional relationships identified from multiple electronic drawings, the construction information from different drawings is associated to establish a multi-view data association relationship for the same bridge component. For the same associated component, geometric parameters, structural parameters, material parameters, and reinforcement parameters from different drawings are integrated to generate structured parameter data for the bridge component. When there are parameter conflicts or missing parameters in different drawings, a knowledge graph reasoning model is constructed based on the bridge structural relationships, bridge information, and calculation rules for each structural component. This model uses the bridge as the entity node, components as attribute nodes, and component relationships as edges. Combined with bridge design specifications and contextual information, the parameter data is validated for consistency and compensated to obtain complete structured component data for the bridge. The reasoning model includes: entity nodes including bridges, bridge spans, piers, cap beams, abutments, pile foundations, beam bodies, and supports; attribute nodes including dimensions, materials, reinforcement, elevation, station number, and mileage information; and relationship edges including connection relationships, support relationships, inclusion relationships, upstream and downstream relationships, and spatial positional relationships, thus forming a structured association network for bridge components.

[0021] Further, in the parameter mapping rule base, component family template information corresponding to the structural type of the component is retrieved to determine the target family file for 3D modeling; then, the parameter name is matched with the family parameters in the modeling system, and the parameter name is mapped to the corresponding family parameter field in the target family file of the modeling through a pre-established parameter lookup table; the corresponding parameter value is bound to the family parameter field to form a component modeling parameter data set containing the constructed parameter value, family name, and family parameter name, for use in subsequent automated modeling steps.

[0022] In existing technologies, the parameter information obtained from image recognition usually needs to be manually converted into parameter fields in modeling software and manually input into the modeling system, resulting in low efficiency and easy data errors. This invention establishes a parameter lookup table to create a mapping relationship between image recognition parameters and family parameters of modeling software, enabling the recognized parameters to be automatically converted into a parameter format that the modeling system can recognize. This achieves automatic data conversion from image parameter recognition to 3D modeling parameter input, improving the automation level of bridge model construction.

[0023] In one implementation, the parameter data is organized into a structured data format, which includes the component family name and the corresponding parameter name and parameter value information. In the final output set, each component modeling parameter data includes the component family name, family parameter name, and parameter value.

[0024] Furthermore, the component modeling parameter data is encapsulated as network request data and a modeling request is sent to the automated modeling service via an HTTP interface. This allows the automated modeling service program, which is encapsulated as a user-free, server-side program, to parse the structured parameter data, locate the corresponding component family file based on the component's family name, silently start the modeling software runtime environment, load the component family file without starting the user interface, batch assign values ​​to the parameters in the component family file based on the component's family parameter name and parameter value, and perform a model update operation to generate the corresponding 3D component model.

[0025] The modeling software's parameter interface is called to obtain the volume information of the component model in order to generate the corresponding concrete volume data of the component.

[0026] Furthermore, the generated 3D model undergoes a lightweighting process, converting the native model format of the modeling software into a 3D model format suitable for online display, and generating a corresponding lightweight model file.

[0027] The final output modeling results include component names, component volume data, and lightweight 3D model data. The lightweight 3D model data includes model file data for web display.

[0028] In existing technologies, the three-dimensional modeling and quantity calculation of bridge components usually require manual input of parameters and manual generation of models in modeling software, and then the concrete volume is calculated through model analysis tools. The whole process relies on manual operation and is inefficient.

[0029] This invention constructs an automated modeling service, deploying modeling software on a server in a user-friendly manner and receiving parameter data via an HTTP interface to achieve parameter-driven automated modeling of components. Simultaneously, after model generation, it automatically acquires the component volume and generates a lightweight 3D model file, automating the entire process from parameter identification and model generation to quantity calculation and 3D visualization. This significantly improves the efficiency of quantity calculation and 3D model generation for bridge structures.

[0030] Furthermore, after reconstructing the component graphic region, component type identification is performed. The component graphic region is merged with its associated parameter annotation region, and the minimum bounding region containing both the component graphic and parameter annotations is calculated. This region is then cropped to obtain the complete component graphic region. The cropped component graphic region is then used to identify the component's structural type and structured parameter data. By jointly reconstructing the component graphic region and parameter annotation region, and combining this with a component classification model for identification, the identification results include not only parameter information but also component type information, providing a complete data foundation for subsequent parameter mapping and automated modeling.

[0031] The automatic component modeling and quantity calculation system based on drawing recognition includes a drawing recognition service module, a parameter processing module, and a BIM automatic modeling service module. It adopts the automatic component modeling and quantity calculation method based on drawing recognition to identify components in drawings and perform automatic component modeling and quantity calculation.

[0032] The drawing recognition service module is used to extract component parameter annotation data from engineering drawings;

[0033] The parameter processing module is used to complete parameter association and parameter mapping;

[0034] The BIM automatic modeling service module is used to automatically generate three-dimensional component models and calculate concrete volume based on the identification parameters.

[0035] The system is deployed in a server environment. The drawing recognition service module receives engineering drawing images and outputs structured parameter data through a web interface, while the BIM automatic modeling service module receives modeling parameters and automatically generates component models through a network interface.

[0036] The advantages and beneficial effects of this invention are as follows:

[0037] This invention addresses the problem that structural parameters in bridge engineering images still require manual identification and input into modeling software in existing technologies. By automatically identifying the component parameter annotations in bridge engineering drawings and mapping the identified parameter information to component family parameters in the building information model, the invention achieves automatic generation of 3D models of bridge components and automatic calculation of concrete quantities through automated modeling services. This improves the efficiency of quantity calculation, reduces the degree of manual intervention, and thus reduces the error rate. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the system structure in an embodiment of the present invention.

[0039] Figure 2 This is a flowchart of the method in an embodiment of the present invention. Detailed Implementation

[0040] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0041] like Figure 1 As shown, this invention proposes an automatic component modeling and quantity calculation system based on drawing recognition, including a drawing recognition service module, a parameter processing module, and a BIM automatic modeling service module. The drawing recognition service module extracts component parameter annotation information from engineering drawings; the parameter processing module performs parameter association and mapping; and the BIM automatic modeling service module automatically generates three-dimensional component models and calculates concrete volume based on the recognized parameters.

[0042] The system is deployed in a server environment. The drawing recognition service receives engineering drawing images and outputs structured parameter data through a web interface, while the BIM automatic modeling service receives modeling parameters and automatically generates component models through a network interface.

[0043] Based on the above system, this invention also proposes a method for automatic component modeling and quantity calculation based on drawing recognition, such as... Figure 2 As shown, it includes the following steps:

[0044] Step S1: Obtain the original project images and annotate them with parameters to construct a parameter annotation dataset, preparing for the identification of bridge parameter samples; specifically including the following steps:

[0045] Step S1.1: Obtain the original project images, which can be 1632×1152px in size, including structural drawings of the bridge and related engineering annotations;

[0046] Step S1.2: Add parameter annotations to the component targets in the original project images;

[0047] Use the Labelme annotation tool to annotate the image and set view parameter label categories: "view" and "viewname" to summarize and distinguish the view source to which the parameter belongs; set label categories for bridge part parameters such as "beam length", "pile diameter", and "pillar width";

[0048] Step S1.3: Generate annotation data for each component target to construct the annotation dataset D. label Each data point contains the coordinates of a bounding box (x, y). i , y i , w i , h i ), where x i ,、yi These represent the x and y coordinates of the top-left corner of the numbered box, respectively. i h i These represent the width and length of the numbering box, respectively.

[0049] By constructing a parameter-annotated dataset, training data is formed, thereby ensuring that subsequent model training can learn the relationship between the parameter features and annotations of bridge parts.

[0050] Step S2: Construct an object detection model, train the object detection model using a parameter-annotated dataset, and deploy the trained model; input the engineering drawing to be identified into the deployed object detection model to extract the engineering annotation data.

[0051] Specifically, the construction of the object detection model includes the following steps:

[0052] Step S2.1: Obtain the labeled dataset D label ;

[0053] Step S2.2: Using the PicoDet lightweight object detection network structure in the PaddlePaddle development framework, combined with the lightweight optimization scheme of PP-YOLOE Lite, construct an object detection model and test it using the labeled dataset D. label Conduct training;

[0054] By combining PicoDet's core algorithm, the Anchor-free frame-of-view, with the GIoULoss loss function, the accuracy of small object detection is optimized. GIoU (Generalized Intersection over Union) is defined as follows:

[0055]

[0056] Where A represents the area of ​​the predicted bounding box, B represents the area of ​​the ground truth bounding box, C represents the area of ​​the smallest closed rectangle between A and B, and IoU (Intersection over Union) represents the intersection-over-union ratio between the predicted bounding box and the ground truth bounding box; GIoU is used to solve the problem that IoU cannot accurately reflect the degree of overlap.

[0057] Step S2.3: After training is complete, export the object detection model in PaddleInference format using the FastDeploy deployment tool and deploy it as a service interface in a Docker container. Finally, output the object detection service interface that can be used for inference.

[0058] The target detection model constructed through step S2 enhances the high-precision target detection capability of bridge structure images compared to existing technologies.

[0059] Specifically, the bridge structure engineering image is input into a pre-trained target detection model to detect the engineering annotation information in the image, identify various parameter annotation targets, and obtain the corresponding bounding box position, annotation category, and confidence information. The annotation targets include parameter number annotations, size parameter annotations, length parameter annotations, and connection relationship annotations, etc. At the same time, the detection results are sorted according to the confidence to obtain a candidate annotation target set, and finally outputs a set of annotation detection results containing annotation category, bounding box position, and confidence information.

[0060] Compared to the traditional method of relying on manual identification of engineering drawing parameters, the deep learning object detection model can automatically identify engineering annotation information, realize automatic positioning of parameter annotations, improve the efficiency of drawing information extraction and reduce manual intervention.

[0061] Step S3: Based on the engineering annotation data, locate the component graphic area in the engineering drawing and identify the component structure type. By recognizing the content of the engineering annotation data, obtain structured parameter data containing parameter names and parameter values. Further, based on the view information, component number information and component spatial position relationship in the engineering drawing, establish the component association relationship between different engineering views. Merge the structured parameter data in multiple views belonging to the same entity component to form structured component data containing a complete set of parameters and geometric constraint relationships.

[0062] Specifically, locating the component graphic region involves using the parameter number annotations in the detected parameter data as reference objects, performing connected component analysis on the original bridge structure image, extracting the largest connected region in the image to determine the component graphic outline, and classifying the annotation information belonging to the same component region according to the spatial positional relationship between the annotation target center point and the component graphic region, thereby identifying the component graphic region corresponding to the parameter number annotation, and finally outputting a set of multiple component graphic regions and their corresponding annotation information.

[0063] Existing drawing recognition technologies typically only identify individual annotations, making it difficult to determine the correspondence between annotations and specific components. This invention, through connected component analysis and spatial relationship judgment, achieves automatic association between parameter annotations and component graphics, improving the structured nature of parameter recognition results.

[0064] Specifically, the association and matching of engineering annotation data and component graphics involves analyzing the spatial relationship of parameter data annotated within the same component area through the component region and its annotation information set. Using the parameter number annotation as a reference point, the distance between it and the dimension annotation and the length annotation is calculated, and the corresponding parameter is determined according to the principle of minimum distance. When the detection result lacks dimension parameters or length parameters, a reference position is constructed by analyzing the spatial arrangement relationship between the existing parameters, and the annotation with the smallest distance from the reference position is selected as a supplementary parameter from the candidate annotation area. Finally, a complete parameter combination containing parameter number, dimension parameter, and length parameter corresponding to the component region is output.

[0065] Compared to traditional image recognition methods that only output independent annotation information, this invention automatically establishes logical relationships between parameters through a spatial relationship matching algorithm and completes the parameters when they are missing, thereby improving the completeness and reliability of the parameter recognition results.

[0066] Specifically, the parameter name and its content are obtained by identifying the parameter content. This involves cropping the image based on the position of the parameter annotation area, performing optical character recognition (OCR) on the cropped image area to identify the text or numerical information in the parameter annotation, binding the identified numerical information with the corresponding parameter annotation category, thereby obtaining structured parameter data containing the parameter name and parameter value, and finally outputting structured parameter data containing the parameter name and corresponding parameter value.

[0067] Compared to the traditional method of manually entering parameters into engineering drawings, this invention uses optical character recognition technology to automatically extract parameter values, thereby achieving automatic digital processing of parameter information, improving data entry efficiency and reducing human error.

[0068] Specifically, after reconstructing the component graphic region, component type identification is performed; the component graphic region is merged with its associated parameter annotation region, the minimum bounding region containing the component graphic and parameter annotation is calculated, and the region is cropped to obtain the complete component graphic region; subsequently, the cropped component graphic region is input into the component classification model for identification to determine the bridge structure type to which the component belongs; finally, structured component data containing component region information, component type, and corresponding parameter information is output.

[0069] By jointly reconstructing the component graphic area and parameter annotation area, and combining it with the component classification model for identification, the identification results include not only parameter information but also component type information, providing a complete data foundation for subsequent parameter mapping and automated modeling.

[0070] Specifically, the cross-drawing association and parameter fusion of bridge components in this embodiment of the invention involves unified management of the structural component data identified from different PDF drawings. Based on component number, axis number, station number, bridge span number, mileage information, component name, and spatial position relationship, component information from different drawings is associated to establish a multi-view association relationship for the same bridge component. For the same associated component, geometric parameters, dimensional parameters, material parameters, and reinforcement parameters from each drawing are fused to form a complete bridge component parameter model. When parameter conflicts or missing parameters exist between different drawings, consistency verification and parameter compensation are performed using an artificial intelligence inference model, bridge design specifications, and contextual information to output complete bridge structural component data.

[0071] For example, in one implementation, for bridge pier components, the system first identifies the dimensional parameters in the transverse, longitudinal, and elevation views, and establishes a cross-sectional relationship based on the pier number, span, and station number. The system then integrates the width parameters identified in the transverse view, the length parameters identified in the longitudinal view, and the height parameters identified in the elevation view to automatically generate a three-dimensional parameter description of the corresponding pier. When the transverse width of the pier is 2m and the longitudinal width is 1.5m, the system automatically identifies the pier as a rectangular pier, not a circular pier, and uses the parameters from all three directions as input data for subsequent automatic modeling and quantity calculations.

[0072] The artificial intelligence reasoning model is a knowledge graph constructed based on the structural relationships of the bridge, bridge information, and calculation rules of various structural components. It mainly includes bridge entity nodes, component attribute nodes, and relationship edges between components. Among them, the entity nodes include bridge components such as bridges, bridge spans, piers, cap beams, abutments, pile foundations, beams, and bearings. The attribute nodes include dimensions, materials, reinforcement, elevation, station number, and mileage information. The relationship edges include connection relationships, support relationships, inclusion relationships, upstream and downstream relationships, and spatial location relationships, thus forming a structured network of associations of bridge components.

[0073] Step S4: Match the structured component data with the parameter mapping rule base to generate component modeling parameter data;

[0074] The structured component data obtained in step S3 above includes component type information, component region information, parameter names and their corresponding parameter values. In a pre-established parameter mapping rule base, component family template information corresponding to the component type information is retrieved to determine the target family file for 3D modeling. Subsequently, the parameter names are matched with family parameters in the modeling system. Specifically, through a pre-established parameter lookup table, the parameter names obtained from image recognition are mapped to the corresponding parameter fields in the family file of the modeling software. After completing the parameter name mapping, the corresponding parameter values ​​are bound to the parameter fields in the target family file to form a component modeling parameter data set for subsequent automated modeling steps.

[0075] In one implementation, the parameter data is organized into a structured data format, which includes the component family name and the corresponding parameter name and parameter value information. In the final output set, each component modeling parameter data includes the component family name, family parameter name, and parameter value.

[0076] In existing technologies, the parameter information obtained from image recognition usually needs to be manually converted into parameter fields in modeling software and manually input into the modeling system, resulting in low efficiency and easy data errors. This invention establishes a parameter lookup table to create a mapping relationship between image recognition parameters and family parameters of modeling software, enabling the recognized parameters to be automatically converted into a parameter format that the modeling system can recognize. This achieves automatic data conversion from image parameter recognition to 3D modeling parameter input, improving the automation level of bridge model construction.

[0077] Step S5: Based on the component modeling parameter data, perform automatic modeling and quantity calculation, and also generate lightweight component models.

[0078] Encapsulate the component modeling parameter data, which includes parameter values, component family names, and family parameter names, in a structured data format, and perform the following steps:

[0079] Step S5.1: Encapsulate the component modeling parameter data into network request data and send a modeling request to the automated modeling service via the HTTP interface.

[0080] Step S5.2: The automated modeling service is deployed on the server and encapsulated as a headless modeling service program. Upon receiving a modeling request, the service program parses the structured parameter data and locates the corresponding component family file based on the component family name contained within it.

[0081] Step S5.3: Subsequently, the modeling software runtime environment is started silently, and the component family file is loaded without starting the user interface.

[0082] Step S5.4: After loading the component family file, assign values ​​to the parameters in the family file in batches according to the received component family parameter names and parameter values, and perform a model update operation to generate the corresponding 3D component model.

[0083] After the model is generated, the parameter interface of the modeling software is called to obtain the volume information of the component model in order to obtain the concrete volume data of the corresponding component.

[0084] Furthermore, the generated 3D model undergoes a lightweighting process, converting the native model format of the modeling software into a 3D model format suitable for online display, and generating a corresponding lightweight model file.

[0085] In one implementation, the lightweight model file is a glTF format model file, and a corresponding geometric data file is generated simultaneously for use in web-based 3D model display.

[0086] The final output modeling results include component names, component volume data, and lightweight 3D model data. The lightweight 3D model data includes model file data for web display.

[0087] In existing technologies, the three-dimensional modeling and quantity calculation of bridge components usually require manual input of parameters and manual generation of models in modeling software, and then the concrete volume is calculated through model analysis tools. The whole process relies on manual operation and is inefficient.

[0088] This invention constructs an automated modeling service, deploying modeling software on a server in a user-friendly manner and receiving parameter data via an HTTP interface to achieve parameter-driven automated modeling of components. Simultaneously, after model generation, it automatically acquires the component volume and generates a lightweight 3D model file, automating the entire process from parameter identification and model generation to quantity calculation and 3D visualization. This significantly improves the efficiency of quantity calculation and 3D model generation for bridge structures.

[0089] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatic component modeling and quantity calculation based on drawing recognition, characterized in that: Obtain the original engineering project drawings and their parameter annotations to construct a parameter annotation dataset; A target detection model is constructed, and the target detection model is trained using the parameter-annotated dataset. The trained target detection model is then used to detect targets on the engineering drawings to be identified, so as to extract the engineering annotation data. Based on engineering annotation data, the component graphic areas in the engineering drawings are located and the component structure type is identified. By recognizing the content of the engineering annotation data, structured parameter data containing parameter names and parameter values ​​is obtained. According to the view information, component information and component spatial position relationship in the engineering drawings, the component association relationship between different engineering views is established. The structured parameter data in multiple views belonging to the same entity component are merged to form structured component data containing a complete set of parameters and geometric constraint relationships. The structured component data is matched with the parameter mapping rule base to generate component modeling parameter data; Based on component modeling parameter data, automatic component modeling and quantity calculation are performed.

2. The method for automatic component modeling and quantity calculation based on drawing recognition according to claim 1, characterized in that: The target detection model is trained using a loss function constructed with the generalized intersection-union ratio (GUOR). The GUOR is calculated by subtracting the minimum closed area of ​​the predicted box and the ground truth box from the intersection of their areas, then comparing the absolute value of the difference with the absolute value of the minimum closed area, and finally subtracting the GUOR from the intersection-union ratio of the predicted box and the ground truth box areas to obtain the GUOR of the predicted box and the ground truth box areas.

3. The method for automatic component modeling and quantity calculation based on drawing recognition according to claim 1, characterized in that: The engineering annotation data includes the bounding box position, annotation category, and confidence information of the parameter annotation targets. The data is sorted by confidence level to obtain a set of candidate parameter annotation targets, and finally, a set of annotation detection results containing annotation category, bounding box position, and confidence information is obtained. The annotation categories in the engineering annotation data include parameter number annotations. Based on the parameter number annotations, the engineering drawings are analyzed for connectivity. The outline of the component graphic is determined by extracting the largest connected region in the drawings. According to the spatial positional relationship between the center point of the annotation target and the component graphic region, the annotation information belonging to the same component region is classified to identify the construction graphic region corresponding to the parameter number annotation, thus obtaining a set of component graphic regions and their corresponding annotation information.

4. The method for automatic component modeling and quantity calculation based on drawing recognition according to claim 3, characterized in that: The annotation categories in the engineering annotation data also include dimension parameter annotations and length parameter annotations. Through the component graphic area and its annotation information set, the spatial relationship analysis of the engineering annotation data in the same component area is performed. Taking the parameter number annotation as a reference point, the distance between it and the dimension annotation and the length annotation is calculated respectively. The corresponding parameters are determined according to the principle of minimum distance. The parameter combination containing the parameter number, dimension parameter and length parameter corresponding to the constructed graphic area is obtained as the component graphic area information.

5. The method for automatic component modeling and quantity calculation based on drawing recognition according to claim 4, characterized in that: In the spatial relationship analysis, when the detection results are missing size parameters or length parameters, a geometric constraint model of the component is established based on the parameter correspondence of the same entity component in different engineering views; when some views are missing size parameters, the missing parameters are derived based on the parameters in the acquired views and the structural constraint relationship of the component, and the parameter consistency check is used to determine whether the fused parameters meet the preset structural constraints.

6. The method for automatic component modeling and quantity calculation based on drawing recognition according to claim 1, characterized in that: Based on the component numbers, bridge span numbers, station numbers, axis relationships and spatial position relationships identified from multiple electronic drawings, the construction information from different drawings is associated to establish a multi-view data association relationship for the same bridge component. For the same component after association, the geometric parameters, structural parameters, material parameters and reinforcement parameters from different drawings are integrated to generate structural parameter data of the bridge component. When there are parameter conflicts or missing parameters in different drawings, a knowledge graph reasoning model is constructed based on the bridge structural relationships, bridge information, and calculation rules of each structural component. The model uses the bridge as the entity node, the components as the attribute node, and the component relationships as the edges. Combined with bridge design specifications and contextual information, the parameter data is checked for consistency and compensated to obtain complete bridge structural component data. The reasoning model includes...

7. The method for automatic component modeling and quantity calculation based on drawing recognition according to claim 1, characterized in that: In the parameter mapping rule base, component family template information corresponding to the structural type of the component is retrieved to determine the target family file for modeling; then, the parameter name is matched with the family parameters in the modeling system, and the parameter name is mapped to the corresponding family parameter field in the target family file of the modeling through a pre-established parameter lookup table; the corresponding parameter value is bound to the family parameter field to form component modeling parameter data containing the constructed parameter value, family name, and family parameter name.

8. The method for automatic component modeling and quantity calculation based on drawing recognition according to claim 7, characterized in that: The component modeling parameter data is encapsulated into request data and a modeling request is sent to the automated modeling service. The automated modeling service, which is encapsulated as a user interface-free service, parses the structured parameter data, locates the corresponding component family file according to the family name of the component, starts the modeling software runtime environment in a silent manner, loads the component family file without starting the user interface, assigns batch values ​​to the parameters in the component family file according to the family parameter name and the parameter value of the component, and performs a model update operation to generate the corresponding component model. The modeling software's parameter interface is called to obtain the volume information of the component model in order to generate the corresponding concrete volume data of the component.

9. The method of claim 1, wherein: After reconstructing the component graphic region, component type identification is performed; the component graphic region is merged with its associated parameter annotation region, the minimum bounding region containing the component graphic and parameter annotation is calculated, and the region is cropped to obtain the complete component graphic region; then the structural type and structured parameter data of the component are identified in the cropped component graphic region.

10. A component automatic modeling and quantity calculation system based on drawing recognition, comprising a drawing recognition service module, a parameter processing module and a BIM automatic modeling service module, characterized in that: The automatic component modeling and quantity calculation method based on drawing recognition as described in any one of claims 1 to 9 is used to identify components in drawings and perform automatic component modeling and quantity calculation. The drawing recognition service module is used to extract component parameter annotation data from engineering drawings; The parameter processing module is used to complete parameter association and parameter mapping; The BIM automatic modeling service module is used to automatically generate component models and calculate concrete volume based on the identification parameters.