A building completion model updating and delivery method based on BIM and computer vision fusion

Abstract

The application relates to a building completion model updating delivery method based on BIM and computer vision fusion. The method comprises the following steps: obtaining building completion site images and point cloud data, and obtaining image-point cloud fusion initial data through denoising, redundancy removal and coordinate registration; aligning corresponding components of a BIM design model with identification data space by extracting component identification data, obtaining a comparison base; extracting and matching actual and designed feature points, and performing deviation determination by using a preset threshold; correcting component features of the BIM design model in batches according to a determination result, and obtaining a preliminarily updated BIM completion model; generating a standardized BIM completion model based on secondary correction and point cloud space precision checking; meanwhile, combining a BIM component list and component deviation data to formulate a delivery checking standard, performing full-dimensional comparison between the standardized model and the initial data, checking the consistency of the BIM completion model and the entity, and obtaining a BIM completion model delivery checking report; and the updating delivery of the building completion model is realized.

Description

Technical Field

[0001] This invention belongs to the field of building digitalization technology, specifically relating to a method for updating and delivering building as-built models that integrates BIM and computer vision. Background Technology

[0002] The following areas still need improvement in the BIM model updating and delivery work during the building completion phase: In the process of digital transformation of building engineering, BIM as-built models have become the core digital carrier connecting the construction and operation and maintenance phases. The accuracy and completeness of their data directly determine the efficiency and quality of subsequent equipment management, space optimization, renovation, and other work. However, the current process of updating and delivering building as-built models still faces many technical bottlenecks, which seriously restricts the full release of the value of building digitalization.

[0003] Traditional BIM as-built model updates often rely on manual on-site measurements and manual model adjustments. This approach not only requires significant manpower but also has extremely low efficiency, making it unsuitable for the completion and delivery needs of large and complex buildings. Furthermore, manual comparison is susceptible to subjective judgment and limitations in the accuracy of measurement tools, leading to discrepancies between the dimensions, locations, and spatial relationships of components in the model and the actual on-site entities. This can result in decision-making errors during subsequent operation and maintenance phases.

[0004] With the application of technologies such as 3D scanning and computer vision, some update methods have begun to attempt to combine image or point cloud data for model correction, but there are still obvious defects: methods that rely solely on image data are easily affected by factors such as changes in ambient lighting, component occlusion, and image resolution, resulting in inaccurate component feature recognition and insufficient spatial positioning accuracy; solutions based solely on point cloud data face the problems of large data redundancy and low processing efficiency, and it is difficult to accurately identify the texture, markings, and other detailed features of components, which cannot meet the needs of refined model updates.

[0005] Furthermore, existing technologies lack a standardized process system: the coordinate registration algorithms between design models and site data have poor compatibility and are prone to registration deviations; feature point extraction and matching lack unified standards, resulting in insufficient consistency in deviation calculation results for different components; deviation judgment thresholds mostly rely on manual setting based on engineering experience, lacking scientific basis; after model correction, there is a lack of a comprehensive and systematic verification mechanism, making it difficult to fully verify the consistency between model parameters and site entities, resulting in inconsistent quality of delivered BIM as-built models, which cannot support the full life cycle management of buildings.

[0006] Most existing methods can only correct individual components one by one, making it difficult to consider the spatial relationships between components. This can easily lead to conflicts between corrected components and surrounding components. Furthermore, their batch processing capabilities are insufficient, failing to meet the actual needs of rapid completion and delivery of large-scale buildings. These problems result in long update and delivery cycles for BIM as-built models, high costs, and difficulty in guaranteeing quality, severely hindering the implementation and promotion of building digitization technology. Summary of the Invention

[0007] To address the aforementioned problems in the existing technology, this invention provides a method for updating and delivering building as-built models that integrates BIM and computer vision; The objective of this invention can be achieved through the following technical solutions: S1: Obtain the first image data and the second point cloud data, and obtain the initial data for image-point cloud fusion through image denoising, point cloud redundancy removal and coordinate registration operations; S2: Spatially align the corresponding components in the BIM design model with the identification data to obtain a comparison base; The actual feature points and design feature points are extracted, and the deviation data of the corresponding components are obtained through feature point matching and deviation calculation. The determination result is obtained by analyzing the deviation data. S3: Based on the judgment result, perform batch corrections on the component features in the BIM design model to obtain a preliminary updated BIM as-built model; By performing secondary registration and identifying the deviation areas of the components, deviation correction and completion data are obtained; The initially updated BIM as-built model is then revised a second time, and the spatial accuracy of the laser point cloud data is verified to obtain a standardized BIM as-built model. S4: Based on the deviation data between the BIM component list and the corresponding components, obtain the delivery verification standard; The standardized BIM as-built model is compared with the initial data in all dimensions to verify the consistency between the model parameters and the on-site entities, and a model delivery verification report is obtained.

[0008] As a preferred technical solution of the present invention, the specific process of acquiring the initial data of image-point cloud fusion includes: acquiring the acquisition boundary and acquisition density of the first image data and the second point cloud data; The first image data is denoised and redundant data of the second point cloud data is removed to obtain denoised image data and redundant point cloud data. Based on the denoised image data and the deredundant point cloud data, coordinate registration is performed using the RANSAC algorithm to obtain the initial data for image-point cloud fusion.

[0009] Specifically, the spatial alignment is performed as follows: based on the component identification data, the spatial coordinate reference points of the components are extracted, the design coordinate reference points of the corresponding components in the BIM design model are retrieved, coordinate calibration is performed using the least squares method, and coordinate calibration parameters are obtained. Based on the coordinate calibration parameters, the corresponding components in the BIM design model are spatially aligned with the on-site component identification data to obtain a comparison base.

[0010] Specifically, the process of extracting actual feature points and design feature points includes: obtaining the comparison area between the on-site component and the corresponding component in the BIM design model based on the comparison base, and defining the feature point extraction range; The SIFT algorithm is used to process the field components in the initial data to extract the actual feature points of the field components; Using the comparison base and the actual feature points, the design parameters of the corresponding components in the BIM design model are retrieved, and the design feature points of the corresponding components are extracted by model slicing. Based on the actual feature points and the designed feature points, feature points are classified and labeled to obtain the classified set of actual feature points and the set of designed feature points.

[0011] Specifically, the process of obtaining the deviation data of the corresponding components includes: performing feature point matching through the KNN matching algorithm to obtain the matching results of the feature points; Based on the matching results, the deviation data of the corresponding feature points are calculated, and the overall deviation value of the corresponding component is calculated by combining the weighted average algorithm to obtain the component deviation data. Classify and statistically analyze the distribution of deviations to obtain complete deviation data for the corresponding components.

[0012] Specifically, the process of obtaining the determination result includes: determining the complete deviation data with the preset deviation threshold of the corresponding component, obtaining the determination result, and identifying the deviation data of the corresponding component based on the determination result.

[0013] Specifically, the process of performing batch corrections includes: calling the BIM model editing interface based on the judgment result and importing the component parameters that need to be updated; Based on the deviation data of the component that needs to be updated, a correction instruction for the corresponding component is generated; Based on the correction instructions, the features of the corresponding components in the BIM design model are batch corrected to obtain a preliminarily corrected component model set. Integrate the spatial relationships of the relevant components to obtain a preliminary updated BIM as-built model.

[0014] Specifically, the process of performing secondary registration and identifying the deviation areas of the components includes: extracting the spatial coordinate system and component outline information based on the initially updated BIM as-built model; The spatial coordinate system and component outline information are registered with the initial data in a second coordinate system to obtain the registered BIM as-built model. The corresponding components are compared with the actual components on site area by area to identify the deviation areas between the BIM as-built model and the actual on site.

[0015] Specifically, the process of obtaining a standardized BIM as-built model includes: performing secondary correction on the corresponding components in the deviation areas of the initially updated BIM as-built model to obtain a BIM as-built model after secondary correction. Spatial verification was performed by comparing the second point cloud data with the initially updated BIM as-built model point by point to obtain verification data. The verification data is compared with the preset completion and delivery standards to obtain the judgment result; A standardized BIM as-built model is obtained based on the determination results.

[0016] Specifically, the process of obtaining the delivery verification standard includes: obtaining the corresponding components and verification indicators for model delivery verification based on the judgment result; By combining the data requirements of the corresponding components in the building completion and delivery standards, the judgment criteria for the verification items are obtained, resulting in a complete BIM as-built model delivery verification standard.

[0017] Specifically, the process of performing the full-dimensional comparison includes: comparing the standardized BIM as-built model with the initial data component by component to obtain the comparison results; Based on the comparison data of the statistical verification items in the comparison results, a full-dimensional comparison table is generated.

[0018] Specifically, the process of verifying the consistency between the model parameters and the on-site entities includes: extracting deviation data during the comparison process, retrieving the first image data and the second point cloud data, performing a second check on the on-site entities, and obtaining the cause of the deviation; Based on the causes of the deviations, standardized delivery standards for BIM as-built models are obtained. Use the consistency verification results to obtain a complete model delivery verification report, and perform consistency verification between the corresponding model parameters and the on-site entities.

[0019] The beneficial effects of this invention are as follows: it clarifies the acquisition boundaries and density, and combines Gaussian filtering, direct-pass filtering and RANSAC algorithms to make the initial data of image-point cloud fusion more accurate and free of redundancy, providing a reliable data foundation for subsequent comparison; it avoids feature extraction deviations caused by data interference, improving the efficiency and accuracy of the overall process; and it achieves precise spatial alignment between the BIM design model and on-site components by calibrating coordinates using the least squares method, ensuring the consistency of the comparison base.

[0020] By utilizing the SIFT algorithm and model slicing technology, actual and design feature points are extracted and classified to provide data support for deviation calculation. The KNN matching algorithm and weighted average algorithm are used to achieve feature point matching and accurately calculate the overall deviation and distribution of components. Deviation judgment and identification are completed by combining preset thresholds. The BIM model editing interface is called to batch correct component features, and the model is integrated by combining spatial relationships to improve the efficiency of initial update and the integrity of the model.

[0021] Secondary registration and comparison, along with area-by-area verification, accurately identify deviation areas. Then, point cloud point-by-point verification and comparison with delivery standards are performed. Dedicated delivery verification standards are developed based on the component list and deviation data. Component-by-component comparison and secondary verification of deviation causes are conducted in all dimensions to ensure that model parameters are consistent with the actual on-site entities. The generated verification report is complete and detailed, ensuring delivery quality. Attached Figure Description

[0022] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0023] Figure 1 This is a flowchart illustrating a method for updating and delivering a building as-built model that integrates BIM and computer vision, according to the present invention. Figure 2 This is a flowchart of the model correction and standardization process in this invention. Detailed Implementation

[0024] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0025] Please see Figure 1-2 A method for updating and delivering a building as-built model that integrates BIM and computer vision includes: S1: acquiring first image data and second point cloud data, and obtaining initial data for image-point cloud fusion through image denoising, point cloud redundancy removal and coordinate registration operations; S2: Spatially align the corresponding components in the BIM design model with the identification data to obtain a comparison base; The actual feature points and design feature points are extracted, and the deviation data of the corresponding components are obtained through feature point matching and deviation calculation. The determination result is obtained by analyzing the deviation data. S3: Based on the judgment result, perform batch corrections on the component features in the BIM design model to obtain a preliminary updated BIM as-built model; By performing secondary registration and identifying the deviation areas of the components, deviation correction and completion data are obtained; The initially updated BIM as-built model is then revised a second time, and the spatial accuracy of the laser point cloud data is verified to obtain a standardized BIM as-built model. S4: Based on the deviation data between the BIM component list and the corresponding components, obtain the delivery verification standard; The standardized BIM as-built model is compared with the initial data in all dimensions to verify the consistency between the model parameters and the on-site entities, and a model delivery verification report is obtained.

[0026] As a preferred technical solution of the present invention, the specific process of acquiring the initial data of image-point cloud fusion includes: acquiring the acquisition boundary and acquisition density of the first image data and the second point cloud data; The first image data is denoised and redundant data of the second point cloud data is removed to obtain denoised image data and redundant point cloud data. Based on the denoised image data and the deredundant point cloud data, coordinate registration is performed using the RANSAC algorithm to obtain the initial data for image-point cloud fusion.

[0027] In this embodiment, based on the architectural construction drawings, structural as-built drawings, and actual building outline of the high-rise office building, the collection boundary is determined to be the entire building space and the component installation area extending from the exterior walls, covering all floors above ground, underground equipment floors, and ancillary component areas. Simultaneously, the collection density is set according to the component type: high-density collection density for main structural components such as concrete frames and shear walls; medium-density collection density for functional components such as electromechanical main pipelines and curtain wall keels; and conventional density collection density for decorative surfaces and small ancillary components. This ensures that the collected data meets the requirements for refined identification while avoiding excessive data redundancy. For the first image data, a Gaussian filtering algorithm is used for denoising to eliminate image interference caused by uneven lighting, equipment imaging noise, and component reflections, while retaining effective visual information such as component texture and markings, resulting in denoised image data. For the second point cloud data, a combination of statistical filtering and direct-pass filtering algorithms is used to first remove outliers and duplicate collection points from the point cloud, and then filter out environmental point clouds and invalid background points that are not part of the building itself, resulting in deredundant point cloud data. The preprocessed denoised image data and deredundant point cloud data are imported into the registration module, and the RANSAC algorithm is used. The specific formula is as follows: Let the spatial transformation model be The model parameters are solved by randomly sampling k groups of samples, and the interior set S is given. in satisfy: , In the formula: ε is the preset distance threshold (registration error tolerance), d(x i , ) represents the Euclidean distance from the sample point to the model, and m represents the total number of samples involved in the registration. When the proportion of interior point set When α is the interior point rate threshold, it is determined to be the optimal transformation matrix M, which is a 4×4 homogeneous transformation matrix in three-dimensional space: , Solve the spatial transformation matrix to complete the coordinate registration of the image and point cloud data in the same world coordinate system, so as to accurately fuse the visual information of the image with the spatial three-dimensional information of the point cloud and obtain the initial data for image-point cloud fusion.

[0028] Specifically, the spatial alignment is performed as follows: based on the component identification data, the spatial coordinate reference points of the components are extracted, the design coordinate reference points of the corresponding components in the BIM design model are retrieved, coordinate calibration is performed using the least squares method, and coordinate calibration parameters are obtained. Based on the coordinate calibration parameters, the corresponding components in the BIM design model are spatially aligned with the on-site component identification data to obtain a comparison base.

[0029] In this embodiment, unique identification data for each component is extracted from the initial image-point cloud fusion data, including the component's model code, QR code, and physical feature identifier. Based on this identification data, the spatial location of each component on site is determined, and unique and representative geometric feature points on the component are selected as spatial coordinate reference points, such as the corner points of concrete columns, the support endpoints of beams, the flange center points of pipelines, and the splicing points of curtain wall keels, ensuring that the reference points are easily identifiable and have stable spatial positions. Simultaneously, design components corresponding one-to-one with the on-site components are retrieved from the BIM design model, and their design coordinate reference points are extracted to form an on-site reference point set and a design reference point set. The two sets of reference points are imported into the coordinate calibration module, and the on-site reference points and design reference points are spatially fitted and calibrated using the least squares method. The calculated coordinate calibration parameters include translation, rotation, and scaling dimensions, which can eliminate coordinate system deviations between on-site data acquisition and design modeling. Based on the coordinate calibration parameters, the overall spatial coordinates of the BIM design model are adjusted so that the corresponding components in the BIM design model are accurately matched with the actual components corresponding to the site component identification data in space. This achieves spatial alignment between the design model and the site entity, forming a unified comparison base for subsequent deviation comparison and feature point extraction, ensuring that the comparison benchmarks of the two are consistent.

[0030] Specifically, the process of extracting actual feature points and design feature points includes: obtaining the comparison area between the on-site component and the corresponding component in the BIM design model based on the comparison base, and defining the feature point extraction range; The SIFT algorithm is used to process the field components in the initial data to extract the actual feature points of the field components; Using the comparison base and the actual feature points, the design parameters of the corresponding components in the BIM design model are retrieved, and the design feature points of the corresponding components are extracted by model slicing. Based on the actual feature points and the designed feature points, feature points are classified and labeled to obtain the classified set of actual feature points and the set of designed feature points.

[0031] In this embodiment, based on the comparison base that has completed spatial alignment, the overlapping comparison area between each on-site component and its corresponding component in the BIM design model is determined. This area covers the main body and key connection parts of the component. Simultaneously, a feature point extraction range is defined within the comparison area to exclude non-feature interference areas on the component surface, ensuring that the extracted feature points are the key geometric and installation features of the component. Within the defined extraction range, the SIFT algorithm is used to detect, describe, and extract feature points from the on-site components in the initial image-point cloud fusion data. This algorithm can effectively identify features such as corners, inflection points, and edge points of components, and possesses scale invariance and rotation invariance, accurately extracting the actual feature points of the on-site components. Based on the spatial correspondence of the comparison base, combined with the spatial location and feature type of the extracted actual feature points, the complete design parameters of the corresponding component are retrieved from the BIM design model. Using model slicing technology, design feature points that match the actual feature points are extracted one-to-one at the corresponding positions and levels of the design model, ensuring a high degree of correspondence between the feature attributes and spatial location of the design feature points and the actual feature points. The extracted actual and design feature points are classified and marked according to component type, floor location, and feature attributes. For example, frame column - 1st floor - C1 - corner point, air duct - 5th floor - F1 - flange point. The marked feature points are then integrated to form a structured and traceable set of classified actual and design feature points.

[0032] Specifically, the process of obtaining the deviation data of the corresponding components includes: performing feature point matching through the KNN matching algorithm to obtain the matching results of the feature points; Based on the matching results, the deviation data of the corresponding feature points are calculated, and the overall deviation value of the corresponding component is calculated by combining the weighted average algorithm to obtain the component deviation data. Classify and statistically analyze the distribution of deviations to obtain complete deviation data for the corresponding components.

[0033] In this embodiment, the classified actual feature point set and the design feature point set are input into the feature point matching module. The KNN matching algorithm is used to perform neighborhood search and similarity calculation on the two sets of feature points. The K-nearest neighbor matching criterion is employed to achieve precise one-to-one matching of feature points, eliminating invalid matching points and ensuring the accuracy and uniqueness of the matching results. Based on the matching results, the deviation values ​​of each matched feature point in the X, Y, and Z coordinate axes are calculated according to the coordinate values ​​in the same world coordinate system, obtaining the three-dimensional deviation data of a single feature point. Differentiated weights are assigned to different feature points according to their importance and weight proportion in the component. For example, the core installation points and stress-bearing feature points of the component have higher weights than ordinary surface feature points. A weighted average algorithm is used to calculate the weighted deviation data of all feature points of a single component, obtaining the overall deviation value of the component, forming the basic component deviation data. Meanwhile, the component deviation data is classified and statistically analyzed in multiple dimensions, including the distribution of deviations along the X / Y / Z axes by deviation direction, the proportion of slight, moderate, and severe deviations by deviation magnitude, and the concentration and dispersion of deviations by feature point location. By integrating the overall deviation value, single feature point deviation data, and deviation distribution statistical results, the complete deviation data of the corresponding component is obtained, which comprehensively reflects the actual deviation of the component.

[0034] Specifically, the process of obtaining the determination result includes: determining the complete deviation data with the preset deviation threshold of the corresponding component, obtaining the determination result, and identifying the deviation data of the corresponding component based on the determination result.

[0035] In this embodiment, the preset component deviation thresholds are formulated based on the unified standards for construction quality acceptance, the design specifications of various building disciplines, and the customized completion and delivery requirements of the project. Different deviation threshold standards are set for different types and functions of components. For example, the deviation threshold for structural components is stricter than that for decorative components, and the deviation threshold for the installation position of electromechanical pipelines is adapted to the on-site construction specifications. The complete deviation data of each component is compared with the corresponding preset deviation threshold item by item and in all dimensions. The focus is on checking whether the deviation of a single feature point exceeds the threshold, whether the overall deviation value of the component meets the acceptance requirements, and whether there is a concentrated deviation. Based on the comparison results, the judgment results are output and divided into three categories: qualified deviation, slightly out of tolerance, and severely out of tolerance. Qualified deviation means that the component meets the completion and delivery requirements, slightly out of tolerance means that the deviation exceeds the threshold but can be compensated by model correction, and severely out of tolerance means that the deviation is too large and the cause needs to be checked in conjunction with the actual site conditions. Based on the judgment results, the deviation data of the corresponding components are differentiated and marked. Qualified deviation data are marked in a regular manner, slightly out-of-tolerance data are marked with a bright yellow mark and associated with the deviation value and location, and severely out-of-tolerance data are marked with a red key mark and the type and concentrated area of ​​the out-of-tolerance. The marked deviation data can be directly connected to the subsequent model correction module to achieve precise correction.

[0036] Specifically, the process of performing batch corrections includes: calling the BIM model editing interface based on the judgment result and importing the component parameters that need to be updated; Based on the deviation data of the component that needs to be updated, a correction instruction for the corresponding component is generated; Based on the correction instructions, the features of the corresponding components in the BIM design model are batch corrected to obtain a preliminarily corrected component model set. Integrate the spatial relationships of the relevant components to obtain a preliminary updated BIM as-built model.

[0037] In this embodiment, based on the deviation determination results, components requiring model updates and corrections are selected. Data exchange is achieved with the BIM design platform via a system interface. The standardized editing interface of the BIM model is called to batch import the original design parameters, identification information, and complete deviation data of the components to be updated into the model correction module. Based on the imported deviation data, personalized correction instructions are automatically generated for each component according to its spatial coordinates, dimensional parameters, geometric features, and installation location. These instructions clearly specify the correction location, correction value, correction method, and spatial adjustment direction, achieving accurate generation and batch distribution of correction instructions. The correction instructions are synchronized to the BIM design model, automatically batch correcting the feature parameters of the corresponding components in the model, replacing the traditional manual adjustment mode and significantly improving correction efficiency. After correction, the preliminarily corrected individual component models are exported and integrated to form a preliminarily corrected component model set. The spatial relationships of the component model set are verified and integrated. The focus is on checking whether the overlapping relationship, connection relationship, and arrangement spacing of the corrected components with surrounding components are reasonable. This eliminates spatial conflicts and positional interference caused by the correction of individual components, ensuring the spatial relevance and coordination of each component. Finally, the components are integrated to form a preliminary updated BIM as-built model.

[0038] Specifically, the process of performing secondary registration and identifying the deviation areas of the components includes: extracting the spatial coordinate system and component outline information based on the initially updated BIM as-built model; The spatial coordinate system and component outline information are registered with the initial data in a second coordinate system to obtain the registered BIM as-built model. The corresponding components are compared with the actual components on site area by area to identify the deviation areas between the BIM as-built model and the actual on site.

[0039] In this embodiment, the overall spatial coordinate system and precise contour information of each component are extracted from the initially updated BIM as-built model, including the 3D shape of the component, boundary features, and connection node contours, ensuring the integrity of the coordinate system and the accuracy of the contour information. The extracted spatial coordinate system and component contour information are then subjected to secondary coordinate registration with the initial data from image-point cloud fusion. Through fine-tuning of the coordinates, minor coordinate system deviations generated during the initial model correction process are eliminated, enabling a higher-precision spatial match between the initially updated BIM as-built model and the on-site collected data, resulting in a registered BIM as-built model. Using the registered BIM as-built model and the initial data from image-point cloud fusion as dual bases, each component undergoes a refined comparison floor-by-floor, area-by-area, and part-by-part. Contour matching, feature point comparison, and spatial position verification are employed to comprehensively compare the model components with the on-site physical components, accurately identifying areas where deviations still exist between the BIM as-built model and the on-site physical components. The specific range, geometry, spatial location, and relative relationship with surrounding components of these deviation areas are clearly defined, and the characteristic information of these deviation areas is recorded.

[0040] Specifically, the process of obtaining a standardized BIM as-built model includes: performing secondary correction on the corresponding components in the deviation areas of the initially updated BIM as-built model to obtain a BIM as-built model after secondary correction. Spatial verification was performed by comparing the second point cloud data with the initially updated BIM as-built model point by point to obtain verification data. The verification data is compared with the preset completion and delivery standards to obtain the judgment result; A standardized BIM as-built model is obtained based on the determination results.

[0041] In this embodiment, for the identified deviation areas, combined with the deviation area identification report and deviation correction and completion data, a secondary fine-tuning correction is performed on the components in the corresponding areas of the initially updated BIM as-built model. The correction process considers both the local features of the components and their overall spatial relationships, performing targeted parameter adjustments, contour optimization, and position calibration on the deviation areas to ensure a high degree of fit between the corrected components and the actual site entities. This results in a second-corrected BIM as-built model. The original second point cloud data is retrieved and compared point-by-point with the second-corrected BIM as-built model. A comprehensive spatial accuracy check is performed on the spatial coordinates, geometric dimensions, positional relationships, contour features, and other parameters of each component in the model. The deviation between the model points and the site point cloud points is recorded point by point, forming complete verification data that comprehensively reflects the spatial accuracy and matching degree of the model. Spatial accuracy verification is used for point-by-point spatial verification of the second-corrected BIM model and point cloud data to verify the consistency between the model and the actual site entities. Let the model point set be M(x... m ,y m ,zm The point cloud set is C(x). c ,y c ,z c The point-by-point verification deviation EMC is: , In the formula: j is the index of the point cloud points adjacent to model point i; η is the proportion of qualified verification points, which is the core indicator for determining whether the model has reached standardization. , In the formula: δ is the indicator function (1 if the condition is met, 0 otherwise), is the verification accuracy threshold, and n is the total number of points participating in the verification; when η≥95% (accuracy requirement for completion and delivery), the model space accuracy is determined to meet the standard, and a standardized BIM as-built model is generated.

[0042] The verification data is compared item by item with the preset building completion and delivery standards and the project-customized accuracy requirements. The focus is on verifying whether the overall accuracy of the model, the key parameters of components, and the matching degree of spatial location meet the completion and delivery requirements. Verification results are output. Models that pass the verification are directly designated as standardized BIM as-built models. For areas with slight deviations, final adjustments are made to ensure all parameters meet the delivery standards, ultimately resulting in a standardized BIM as-built model that meets the requirements of the entire building completion and delivery process.

[0043] Specifically, the process of obtaining the delivery verification standard includes: obtaining the corresponding components and verification indicators for model delivery verification based on the judgment result; By combining the data requirements of the corresponding components in the building completion and delivery standards, the judgment criteria for the verification items are obtained, resulting in a complete BIM as-built model delivery verification standard.

[0044] In this embodiment, based on the preliminary deviation judgment results, the verification results after model correction, and the component composition of the standardized BIM as-built model, the scope of components participating in the BIM as-built model delivery verification is determined, covering all professional types of components such as building structure, MEP, curtain wall, and decoration, ensuring no omissions and full coverage. For the core attributes and as-built delivery requirements of each component, corresponding verification indicators are formulated, including spatial location accuracy, dimensional parameter accuracy, geometric feature matching degree, spatial correlation between components, and integrity of identification information, ensuring that the verification indicators cover the core quality dimensions of the model. Combining national and industry-issued building as-built delivery standards, as well as the data requirements for corresponding components stipulated in project design documents and construction acceptance requirements, specific and quantifiable judgment standards are formulated for each verification indicator, clarifying the qualified range, allowable deviation threshold, verification method, data collection requirements, and judgment rules for each indicator. The verification indicators and corresponding judgment standards for each component are systematically integrated, hierarchically organized according to component professional type, verification dimension, and importance level, forming a complete BIM as-built model delivery verification standard that is clearly hierarchical, requires specific details, is implementable, and quantifiable.

[0045] Specifically, the process of performing the full-dimensional comparison includes: comparing the standardized BIM as-built model with the initial data component by component to obtain the comparison results; Based on the comparison data of the statistical verification items in the comparison results, a full-dimensional comparison table is generated.

[0046] In this embodiment, based on the established BIM as-built model delivery verification standards, the full-dimensional comparison scope is determined to include all components and verification indicators of the standardized BIM as-built model, ensuring full coverage of components and no omissions of indicators. Simultaneously, suitable comparison methods are determined, comprehensively employing multiple methods such as coordinate comparison, contour comparison, feature point comparison, parameter comparison, and identification information comparison. The optimal comparison method is selected for different verification indicators to ensure the accuracy of the comparison results. According to the comparison scope and methods, the standardized BIM as-built model and the initial data from image-point cloud fusion are subjected to a detailed, component-by-component and indicator-by-indicator full-dimensional comparison. The matching status and deviation data for each component and each verification indicator are recorded in detail, clarifying the correspondence between design values, model values, and site values, and outputting the specific comparison results for each component and each indicator. The comparison data of all verification items are classified, statistically analyzed, and summarized. They are presented in a structured manner according to the professional type of components, floor location, and verification index type, generating a full-dimensional comparison table containing core contents such as comparison items, design standard values, actual values ​​of the model, on-site collected values, deviation values, and deviation judgment results, presenting the overall matching status between the standardized BIM as-built model and the on-site entity.

[0047] Specifically, the process of verifying the consistency between the model parameters and the on-site entities includes: extracting deviation data during the comparison process, retrieving the first image data and the second point cloud data, performing a second check on the on-site entities, and obtaining the cause of the deviation; Based on the causes of the deviations, standardized delivery standards for BIM as-built models are obtained. Use the consistency verification results to obtain a complete model delivery verification report, and perform consistency verification between the corresponding model parameters and the on-site entities.

[0048] In this embodiment, all comparison data with deviations are extracted from the full-dimensional comparison table, including key information such as the name of the deviating component, the deviation verification index, the deviation value, and the deviation type. Based on this deviation data, the corresponding first image data and second point cloud data are accurately retrieved. Combined with on-site verification and construction data tracing, a comprehensive analysis of the specific causes of the deviations is conducted. The causes of deviations are categorized into different types, such as construction errors, data acquisition errors, design optimization adjustments, and changes in on-site working conditions, and the specific circumstances and impact range of each cause are recorded. Based on the causes of deviations and the actual completion and delivery requirements of the project, the delivery standards of the standardized BIM as-built model are reasonably optimized and improved, clarifying the acceptable range of deviations, rectification requirements, and exemption conditions, so that the delivery standards are more in line with the actual on-site conditions. By integrating all information such as the results of multi-dimensional comparisons, deviation cause analysis, delivery standard optimization results, and spatial accuracy verification results, a comprehensive and systematic verification and judgment is made on the consistency of various parameters of the standardized BIM as-built model with the on-site entities. Finally, a complete model delivery verification report is generated, which includes core contents such as overall model quality evaluation, deviation details and cause analysis, rectification suggestions and handling solutions, verification conclusions and delivery opinions. The consistency verification of all parameters of the model with the on-site entities is completed through this report.

[0049] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for updating and delivering building as-built models that integrates BIM and computer vision, characterized in that, include: S1: Obtain the first image data and the second point cloud data, and obtain the initial data for image-point cloud fusion through image denoising, point cloud redundancy removal and coordinate registration operations; S2: Spatially align the corresponding components in the BIM design model with the identification data to obtain a comparison base; The actual feature points and design feature points are extracted, and the deviation data of the corresponding components are obtained through feature point matching and deviation calculation. The determination result is obtained by analyzing the deviation data. S3: Based on the judgment result, perform batch corrections on the component features in the BIM design model to obtain a preliminary updated BIM as-built model; By performing secondary registration and identifying the deviation areas of the components, deviation correction and completion data are obtained; The initially updated BIM as-built model is then revised a second time, and the spatial accuracy of the laser point cloud data is verified to obtain a standardized BIM as-built model. S4: Based on the deviation data between the BIM component list and the corresponding components, obtain the delivery verification standard; The standardized BIM as-built model is compared with the initial data in all dimensions to verify the consistency between the model parameters and the on-site entities, and a model delivery verification report is obtained.

2. The method according to claim 1, characterized in that, The specific process of acquiring the initial data for image-point cloud fusion includes: acquiring the acquisition boundary and acquisition density of the first image data and the second point cloud data; The first image data is denoised and redundant data of the second point cloud data is removed to obtain denoised image data and redundant point cloud data. Based on the denoised image data and the deredundant point cloud data, coordinate registration is performed using the RANSAC algorithm to obtain the initial data for image-point cloud fusion.

3. The method according to claim 1, characterized in that, The spatial alignment process involves: extracting the spatial coordinate reference points of the components based on the component identification data, retrieving the design coordinate reference points of the corresponding components in the BIM design model, performing coordinate calibration using the least squares method, and obtaining coordinate calibration parameters. Based on the coordinate calibration parameters, the corresponding components in the BIM design model are spatially aligned with the on-site component identification data to obtain a comparison base.

4. The method according to claim 1, characterized in that, The specific process of extracting actual feature points and design feature points includes: obtaining the comparison area between the on-site components and the corresponding components of the BIM design model based on the comparison base, and defining the feature point extraction range; The SIFT algorithm is used to process the field components in the initial data to extract the actual feature points of the field components; Using the comparison base and the actual feature points, the design parameters of the corresponding components in the BIM design model are retrieved, and the design feature points of the corresponding components are extracted by model slicing. Based on the actual feature points and the designed feature points, feature points are classified and labeled to obtain the classified set of actual feature points and the set of designed feature points.

5. The method according to claim 1, characterized in that, The specific process of obtaining the deviation data of the corresponding components includes: performing feature point matching through the KNN matching algorithm to obtain the matching results of the feature points; Based on the matching results, the deviation data of the corresponding feature points are calculated, and the overall deviation value of the corresponding component is calculated by combining the weighted average algorithm to obtain the component deviation data. Classify and statistically analyze the distribution of deviations to obtain complete deviation data for the corresponding components.

6. The method according to claim 1, characterized in that, The specific process of obtaining the judgment result includes: judging the complete deviation data with the preset deviation threshold of the corresponding component, obtaining the judgment result, and marking the deviation data of the corresponding component based on the judgment result.

7. The method according to claim 1, characterized in that, The specific process for batch correction includes: calling the BIM model editing interface based on the judgment result and importing the component parameters that need to be updated; Based on the deviation data of the component that needs to be updated, a correction instruction for the corresponding component is generated; Based on the correction instructions, the features of the corresponding components in the BIM design model are batch corrected to obtain a preliminarily corrected component model set. Integrate the spatial relationships of the relevant components to obtain a preliminary updated BIM as-built model.

8. The method according to claim 1, characterized in that, The specific process of performing secondary registration and identifying the deviation area of ​​the component includes: extracting the spatial coordinate system and component outline information based on the initially updated BIM as-built model; The spatial coordinate system and component outline information are registered with the initial data in a second coordinate system to obtain the registered BIM as-built model. The corresponding components are compared with the actual components on site area by area to identify the deviation areas between the BIM as-built model and the actual on site.

9. The method according to claim 1, characterized in that, The specific process of obtaining a standardized BIM as-built model includes: performing secondary correction on the corresponding components in the deviation areas of the initially updated BIM as-built model to obtain a BIM as-built model after secondary correction. Spatial verification was performed by comparing the second point cloud data with the initially updated BIM as-built model point by point to obtain verification data. The verification data is compared with the preset completion and delivery standards to obtain the judgment result; A standardized BIM as-built model is obtained based on the determination results.

10. The method according to claim 1, characterized in that, The specific process of obtaining the delivery verification standard includes: obtaining the corresponding components and verification indicators for model delivery verification based on the judgment result; By combining the data requirements of the corresponding components in the building completion and delivery standards, the judgment criteria for the verification items are obtained, resulting in a complete BIM as-built model delivery verification standard.

11. The method according to claim 1, characterized in that, The specific process of performing the full-dimensional comparison includes: comparing the standardized BIM as-built model with the initial data component by component to obtain the comparison results; Based on the comparison data of the statistical verification items in the comparison results, a full-dimensional comparison table is generated.

12. The method according to claim 1, characterized in that, The specific process of verifying the consistency between the model parameters and the on-site entities includes: extracting deviation data during the comparison process, retrieving the first image data and the second point cloud data, performing a second verification on the on-site entities, and obtaining the cause of the deviation; Based on the causes of the deviations, standardized delivery standards for BIM as-built models are obtained. Use the consistency verification results to obtain a complete model delivery verification report, and perform consistency verification between the corresponding model parameters and the on-site entities.

Images