Fabricated building quality monitoring system based on BIM and point cloud integration

Abstract

The invention provides a fabricated building quality monitoring system based on BIM and point cloud integration, and relates to the technical field of data processing, and the system comprises a BIM model feature point obtaining module which is used for obtaining BIM model feature points; the point cloud feature point obtaining module is used for obtaining point cloud feature points; the preliminary matching result obtaining module is used for carrying out descriptor matching and matching pair screening; the optimization matching result obtaining module is used for performing evaluation transformation and error calculation; the fine alignment result obtaining module is used for carrying out preliminary transformation calculation, nearest neighbor searching and error minimization; and the quality improvement suggestion generation module is used for carrying out error analysis and manual correction and carrying out monitoring and quality tracing. According to the method and the device, the technical problem of insufficient feature point extraction and matching precision in the prior art can be solved, and the technical effect of improving the alignment precision of the point cloud data and the BIM model by optimizing the feature point extraction and matching precision is achieved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to an assembled building quality monitoring system based on BIM and point cloud integration. Background Art

[0002] In the quality monitoring of prefabricated buildings, the alignment technology between point cloud data and BIM models is crucial, but existing technologies face multiple challenges in this process.

[0003] At present, the existing methods have deficiencies in feature point extraction, matching and error correction, which makes it difficult for the accuracy of the final alignment results to meet high-precision requirements. The algorithm complexity and large amount of calculation in the integration process make it difficult to ensure real-time performance and efficiency. In addition, the existing methods lack effective mechanisms when dealing with noise and outliers, which further reduces the overall alignment accuracy. Therefore, an improved method is urgently needed to improve the accuracy of alignment and mapping between point cloud data and BIM models by optimizing feature point extraction, improving the robustness of matching algorithms and strengthening error correction.

[0004] In summary, there are technical problems in the prior art due to insufficient accuracy of feature point extraction and matching algorithms and imperfect error correction mechanisms, which lead to low accuracy in alignment and mapping of point cloud data with BIM models, further affecting the accuracy and reliability of the alignment results, thereby limiting its effectiveness in high-precision application scenarios. Summary of the invention

[0005] The purpose of this application is to provide a prefabricated building quality monitoring system based on BIM and point cloud integration, so as to solve the technical problems in the prior art that the accuracy of point cloud data and BIM model alignment and mapping is low due to insufficient accuracy of feature point extraction and matching algorithms and imperfect error correction mechanism, which further affects the accuracy and reliability of the alignment results, thereby limiting its effectiveness in high-precision application scenarios.

[0006] In view of the above problems, the present application provides an assembled building quality monitoring system based on BIM and point cloud integration.

[0007] The present application provides a prefabricated building quality monitoring system based on BIM and point cloud integration, including: a BIM model feature point acquisition module, used to acquire a target BIM model according to a target prefabricated building, extract geometric features from the target BIM model, and obtain BIM model feature points; a point cloud feature point acquisition module, used to acquire target point cloud data according to the target prefabricated building, extract point cloud features from the target point cloud data, and obtain point cloud feature points; a preliminary matching result acquisition module, used to perform descriptor matching based on the BIM model feature points and the point cloud feature points, obtain matching pairs, and perform matching pair screening , obtain preliminary matching results; an optimized matching result obtaining module, used to evaluate and transform the preliminary matching results, and perform error calculation to obtain optimized matching results; a fine alignment result obtaining module, used to perform preliminary transformation calculation on the optimized matching results, obtain rough alignment results, perform nearest neighbor search and minimize errors on the rough alignment results, and obtain fine alignment results; a quality improvement suggestion generation module, used to perform error analysis and manual correction on the fine alignment results, obtain correction results, monitor and generate quality monitoring reports based on the integrated full process, and perform quality traceability through the quality monitoring reports to generate quality improvement suggestions.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: The BIM model feature point acquisition module is used to acquire the target BIM model according to the target prefabricated building, extract geometric features from the target BIM model, and obtain BIM model feature points; the point cloud feature point acquisition module is used to acquire the target point cloud data according to the target prefabricated building, extract point cloud features from the target point cloud data, and obtain point cloud feature points; the preliminary matching result acquisition module is used to perform descriptor matching based on the BIM model feature points and the point cloud feature points to obtain matching pairs, and perform matching pair screening to obtain preliminary matching results; the optimized matching result acquisition module is used to evaluate and transform the preliminary matching results, and perform error calculation to obtain The module is used to perform preliminary transformation calculation on the optimized matching result to obtain a rough alignment result, perform nearest neighbor search and minimize the error on the rough alignment result to obtain a fine alignment result; the module is used to perform error analysis and manual correction on the fine alignment result to obtain the correction result, monitor and generate a quality monitoring report based on the integrated whole process, and perform quality traceability through the quality monitoring report to generate quality improvement suggestions, thereby achieving the technical goals of optimizing feature point extraction, improving the robustness of the matching algorithm and strengthening error correction, and achieving the technical effect of significantly improving the alignment and mapping accuracy of point cloud data and BIM models.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0011] Figure 1 This is a schematic diagram of the structure of the prefabricated building quality monitoring system based on BIM and point cloud integration in this application; Figure 2 This is a flow chart of the BIM model feature point acquisition module in the prefabricated building quality monitoring system based on BIM and point cloud integration in this application.

[0012] Description of reference numerals: BIM model feature point acquisition module 11, point cloud feature point acquisition module 12, preliminary matching result acquisition module 13, optimized matching result acquisition module 14, fine alignment result acquisition module 15, quality improvement suggestion generation module 16. DETAILED DESCRIPTION

[0013] This application provides a prefabricated building quality monitoring system based on BIM and point cloud integration, which solves the technical problem in the prior art that the accuracy of alignment and mapping of point cloud data and BIM models is low due to insufficient accuracy of feature point extraction and matching algorithms and imperfect error correction mechanisms, which further affects the accuracy and reliability of the alignment results, thereby limiting its effectiveness in high-precision application scenarios. The technical goals of optimizing feature point extraction, improving the robustness of matching algorithms, and strengthening error correction are achieved, achieving the technical effect of significantly improving the accuracy of alignment and mapping of point cloud data and BIM models.

[0014] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0015] For examples, please see the attached Figure 1 , this application provides a prefabricated building quality monitoring system based on BIM and point cloud integration, specifically including: The BIM model feature point acquisition module 11 is used to acquire a target BIM model according to a target prefabricated building, extract geometric features from the target BIM model, and acquire BIM model feature points.

[0016] Specifically, the target BIM model of the target prefabricated building is obtained through the three-dimensional digital model generated by the architectural design software, and the target BIM model is converted into feature points that can be analyzed. Various algorithms and tools, such as corner detection algorithms, edge detection algorithms and geometric analysis methods, are used to extract key geometric features from the BIM model, including geometric features such as corners, edges and plane intersections, to obtain BIM model feature points, ensuring that the geometric structure of the target BIM model can be accurately represented.

[0017] The point cloud feature point acquisition module 12 is used to acquire target point cloud data according to the target prefabricated building, extract point cloud features from the target point cloud data, and obtain point cloud feature points.

[0018] Specifically, a 3D laser scanner or photogrammetry technique is used to obtain the target point cloud data of the target prefabricated building. The target point cloud data consists of a large number of 3D coordinate points, which accurately record the geometric shape and structural features of the building. During the acquisition process, the accuracy and stability of the scanning equipment are ensured, and all details of the building are covered as much as possible to obtain complete and high-quality point cloud data. It is a key step to convert point cloud data into feature points that can be used for analysis. First, the point cloud data needs to be preprocessed, including denoising and downsampling, to improve data quality and processing efficiency. Then, the normal vector of each point is determined by neighborhood construction and normal vector calculation. The neighborhood construction can adopt the K nearest neighbor or radius-based method. The normal vector reflects the local surface characteristics of the point and is the basis for the subsequent calculation of curvature and identification of feature points. The curvature value is obtained by calculating the rate of change of the normal vector of each point. The larger the curvature value, the more drastic the surface change at the point, and the point cloud feature point is obtained.

[0019] The preliminary matching result obtaining module 13 is used to perform descriptor matching according to the BIM model feature points and the point cloud feature points to obtain matching pairs, and to screen the matching pairs to obtain preliminary matching results.

[0020] Specifically, descriptor matching is performed between the feature points in the BIM model and the feature points in the point cloud data. Each feature point has a descriptor that can encode the neighborhood geometric characteristics of the point into a numerical form. The similarity of each pair of feature points is determined by calculating the distance between the descriptors of the BIM model feature points and the point cloud feature points. The smaller the distance, the more similar the geometric forms of the two feature points are. All possible feature points are paired, and the similarity of each pair of feature points is calculated to form a list of matching pairs. The matching pairs are screened to filter out unreasonable matching pairs and retain matching pairs that truly have high similarity. Usually a similarity threshold is set, and only matching pairs whose descriptor distance is lower than the similarity threshold will be retained. By comparing the descriptor distance of each pair of feature points, if the distance exceeds the threshold, the matching pair feature points are considered to be geometrically mismatched and are thus eliminated. After screening, the remaining matching pairs form a preliminary matching result.

[0021] The optimized matching result obtaining module 14 is used to evaluate and transform the preliminary matching result and perform error calculation to obtain the optimized matching result.

[0022] Specifically, all matching pairs are extracted from the preliminary matching results, and the matching pairs include point cloud feature points and their corresponding BIM model feature points. The rigid transformation matrix of the matching pairs is calculated, which consists of rotation and translation, and is intended to align the point cloud feature points to the BIM model feature points. The transformation matrix calculation is based on the coordinates of the point cloud feature points and the BIM model feature points, and the optimal rotation and translation parameters are determined by minimizing the distance error. Once the transformation matrix is ​​calculated, it is applied to the point cloud feature points to obtain the transformed point cloud feature point positions. The Euclidean distance between each pair of transformed point cloud feature points and their corresponding BIM model feature points is calculated as the error value, which reflects the degree of alignment between the transformed point cloud feature points and the BIM model feature points. All error values ​​are summarized and analyzed to obtain the optimized matching results, and the evaluation results of the overall transformation effect can be obtained.

[0023] The fine alignment result obtaining module 15 is used to perform preliminary transformation calculation on the optimized matching result to obtain a rough alignment result, perform nearest neighbor search and minimize error on the rough alignment result to obtain a fine alignment result.

[0024] Specifically, all feature point matching pairs are extracted from the optimized matching results, including point cloud feature points and BIM model feature points. Next, the centroid of the feature points, i.e., the average position, is calculated. After the centroid is calculated, the coordinates of all feature points are subtracted from the coordinates of the centroid to perform a centroid removal operation to eliminate the translation effect. Then, the covariance matrix after centroid removal is constructed, and the rotation matrix is ​​obtained by singular value decomposition (SVD). Combining the rotation matrix and the translation vector, the preliminary rigid transformation matrix is ​​calculated, which is applied to the point cloud feature points to obtain the transformed coordinates of the point cloud feature points to form a rough alignment result. A spatial index is established for the point cloud feature points after rough alignment to accelerate the search process of the nearest neighbor point. For each transformed point cloud feature point, the nearest neighbor point is found in the BIM model feature point through the spatial index, that is, the BIM model feature point with the smallest distance to the point cloud feature point. Next, the Euclidean distance between each point cloud feature point and its nearest neighbor BIM model feature point is calculated to represent the error in the current alignment state. In order to minimize the error, the rigid transformation matrix needs to be iteratively updated. Until the error converges to the preset threshold or the maximum number of iterations is reached. Each iteration updates the transformation matrix so that the feature points of the point cloud are more accurately aligned with the feature points of the BIM model, and finally a fine alignment result is obtained.

[0025] The quality improvement suggestion generation module 16 is used to perform error analysis and manual correction on the fine alignment results, obtain correction results, monitor and generate a quality monitoring report based on the integrated full process, and perform quality traceability through the quality monitoring report to generate quality improvement suggestions.

[0026] Specifically, all matching pairs are extracted from the fine alignment results, including point cloud feature points and BIM model feature points. The alignment accuracy is evaluated by calculating the residuals between each pair of matching points. All residuals are statistically analyzed, and statistical indicators such as average residual, maximum residual, minimum residual and standard deviation are calculated to form an error analysis report. Then, the alignment results are displayed through visualization tools, and professionals check and manually adjust the feature points for manual correction. The adjustment information and error changes of each correction are recorded, and the correction results are finally obtained. The error analysis and manual correction process are incorporated into an integrated full-process monitoring system to record each step of operation and data in real time, including error calculation, statistical analysis, visualization test and manual correction. The quality monitoring report is automatically generated through the integrated monitoring system. The report content includes the results of each error calculation, the error changes before and after correction, the detailed record of each manual adjustment and its impact on the alignment accuracy, etc. Through comprehensive recording and analysis, the quality monitoring report provides a detailed evaluation and traceability of the entire alignment process. Quality traceability is performed based on the generated quality monitoring report. By analyzing the error sources and distribution in the report and the effect of each manual correction, the main problems and weak links in the alignment process are identified. Summarize the types and causes of errors that are prone to occur in specific scenarios or steps. For example, large matching errors of certain feature points may be due to initial data quality issues, or the algorithms of certain steps need to be optimized. Based on the analysis results, make targeted quality improvement suggestions. Improvement suggestions may include optimizing the point cloud feature point extraction algorithm, improving the error calculation method, enhancing the function of manual correction tools, or strengthening the monitoring of certain key steps.

[0027] The prefabricated building quality monitoring system based on BIM and point cloud integration achieves the technical goals of optimizing feature point extraction, improving the robustness of the matching algorithm and strengthening error correction, and achieves the technical effect of significantly improving the alignment and mapping accuracy of point cloud data and BIM models.

[0028] Further, if Figure 2 As shown, this application also includes: Corner point extraction, edge extraction and plane intersection line extraction are performed according to the target BIM model to obtain geometric feature points; the geometric feature points are described according to feature descriptors, and neighborhood calculation of the geometric descriptors is performed to obtain BIM neighborhood feature distribution, the geometric feature points are marked in the target BIM model, and the BIM model feature points are obtained in combination with the BIM neighborhood feature distribution.

[0029] Specifically, corner point extraction, edge extraction and plane intersection line extraction are performed according to the target BIM model to obtain geometric feature points. Corner point extraction is to identify those points in the model with large curvature changes. Corner points usually appear at the corners of buildings or key nodes of structures. Corner points can be identified by calculating curvature or using corner point detection algorithms. Edge extraction refers to identifying the edge lines of each component in the model. Edge lines are usually the junctions of different components or the locations of surface mutations. Edge extraction can be achieved through edge detection algorithms, such as Canny edge detection. Plane intersection line extraction is to identify the intersection lines between different planes. Intersection lines usually represent interfaces or connections between different components in a building. These intersection lines can be found through geometric analysis. The geometric feature points extracted by the three methods can fully reflect the geometric structural characteristics in the BIM model.

[0030] Next, after obtaining the geometric feature points, the feature descriptors of the geometric feature points are calculated. Feature descriptors are vectors used to describe the local geometric shapes around feature points. Common descriptors include FPFH and SHOT. Geometric descriptors are generated by counting the geometric information in the neighborhood of feature points. After generating the descriptors, it is necessary to perform neighborhood calculations on the geometric descriptors to calculate the distribution of each feature point in the neighborhood and obtain the BIM neighborhood feature distribution, which not only helps to understand the local geometric structure of the feature points, but also improves the accuracy of feature point matching.

[0031] Then, after the feature descriptors and neighborhood calculations are completed, the geometric feature points are marked in the target BIM model, usually by using specific tools in the BIM software to record the location information and descriptor information of each geometric feature point. At the same time, combining the neighborhood feature distribution can further analyze the specific location and significance of the geometric feature points in the model. By marking the feature points, basic data can be provided for subsequent model analysis, alignment and other applications. Combined with the information of the neighborhood feature distribution, the surrounding environment of the feature points can be more comprehensively understood, thereby improving the recognition and application of model feature points.

[0032] Through methods such as corner point extraction, edge extraction, and plane intersection extraction, the geometric feature points in the BIM model are obtained, and then the geometric feature points are described using feature descriptors, and neighborhood calculations are performed to obtain the BIM neighborhood feature distribution. Next, the geometric feature points are marked in the BIM model, and combined with the information of the neighborhood feature distribution, more accurate and comprehensive BIM model feature points are obtained, which not only ensures the accurate extraction and description of the feature points, but also lays a solid foundation for subsequent model processing and application. Through the systematic feature extraction and description process, the efficient analysis and application of the BIM model can be better realized.

[0033] Furthermore, the present application also includes: A first point cloud data point is extracted according to the target point cloud data; a neighborhood is constructed and a normal vector is calculated for the first point cloud data point to obtain a first normal vector; a curvature is calculated according to a rate of change of the first normal vector to obtain a first curvature value; a distance change from a first neighborhood is calculated according to the first point cloud data point, and if the rate of change exceeds a preset rate of change threshold, an edge point judgment result is obtained; the first curvature value is screened according to the curvature threshold, and a first feature point is obtained in combination with the edge point judgment result; a neighborhood calculation of a point cloud descriptor is performed according to the first feature point to obtain a first point cloud neighborhood feature distribution, the point cloud feature point is marked in the target point cloud data, and a first point cloud feature point is obtained in combination with the first point cloud neighborhood feature distribution, and the first point cloud feature point is added to the point cloud feature point.

[0034] Specifically, a point is selected from the target point cloud data as the first point cloud data point. The selection of the first point cloud data point can be random or based on a preset criterion, such as the density or position of the point, and then a reference point is determined to start detailed analysis and processing.

[0035] Then, after the first point cloud data point is determined, its neighborhood is constructed for the first point cloud data point. The neighborhood construction usually adopts K-nearest neighbor (KNN) or radius-based methods, and a certain number or range of points around the first point cloud data point are regarded as its neighborhood. Then, the normal vector of the neighborhood point, i.e., the first normal vector, is calculated using algorithms such as principal component analysis (PCA). The normal vector reflects the local surface characteristics of the point and is the basis for the subsequent calculation of curvature and other geometric features.

[0036] Next, after calculating the normal vector, the curvature is calculated by analyzing the change of the first normal vector in the neighborhood. Specifically, by comparing the normal vector directions of the first point cloud data point and its neighborhood points, the rate of change of the normal vector is calculated to obtain the first curvature value. The larger the curvature value, the more drastic the surface change at that point.

[0037] Next, in addition to the change in the normal vector, the change in the distance between the point and the neighboring points also needs to be considered. By calculating the change in the distance between the first point cloud data point and its neighboring points, it is determined whether its rate of change exceeds the preset rate of change threshold. If it exceeds, it is considered that the point may be an edge point, and the edge point judgment result is obtained. Edge points are usually located at the boundaries or mutations of geometric shapes, and are an important basis for feature point recognition.

[0038] Then, in order to further determine the feature point, it is necessary to combine the curvature value and the edge point judgment result. First, the first curvature value is screened out according to the preset curvature threshold. If the first curvature value is higher than the curvature threshold, it is considered that the point may be a feature point. Then, this result is combined with the edge point judgment result. If the point is a high curvature point or an edge point, it is determined to be the first feature point, ensuring the accuracy and reliability of the feature point.

[0039] Then, after determining the first feature point, its descriptor needs to be calculated. The descriptor is used to describe the local geometric characteristics of the feature point, and is usually generated by calculating the geometric distribution in the neighborhood of the feature point. The specific steps include selecting a suitable descriptor algorithm (such as FPFH or SHOT), and calculating the descriptor based on the neighborhood data of the first feature point, thereby obtaining the neighborhood feature distribution of the first point cloud.

[0040] Then, after the descriptor calculation is completed, the feature points need to be marked in the target point cloud data. The location information and descriptor information of the first feature point are recorded to form the mark of the point cloud feature point. At the same time, the first point cloud neighborhood feature distribution is combined to obtain the first point cloud feature point, so as to facilitate further analysis and processing of the feature point.

[0041] Then, the first point cloud feature point is added to the point cloud feature point, ensuring that all extracted feature points are systematically recorded and saved, providing a complete data basis for subsequent point cloud data processing, alignment and analysis. Through the systematic feature point extraction and marking process, efficient processing and accurate analysis of point cloud data can be achieved.

[0042] By extracting the first point cloud data point from the target point cloud data, constructing the neighborhood and calculating the normal vector, obtaining the curvature value and edge point judgment results, screening out the feature points and calculating the descriptors, and then marking the feature points and adding them to the collection, the accurate extraction and comprehensive description of the feature points of the point cloud data are ensured, providing a solid foundation for subsequent data processing and analysis.

[0043] Furthermore, the present application also includes: Acquire all point cloud data points within a preset point cloud radius of the first point cloud data point to generate a first neighborhood of the first point cloud data point; acquire the nearest K point cloud data points of the first point cloud data point to generate a second neighborhood of the first point cloud data point; perform weighted calculation on the first neighborhood and the second neighborhood based on the coefficient of variation method to obtain the neighborhood of the first point cloud data point, perform neighborhood search, and obtain a neighborhood set of the first point cloud data point; calculate the covariance matrix of the neighborhood set, perform eigenvalue decomposition to obtain eigenvalues ​​and eigenvectors, and obtain the first normal vector.

[0044] Specifically, a predetermined radius is set with the first point cloud data point as the center, and all point cloud data points within the predetermined radius are searched. All point cloud data points contained within the predetermined radius constitute a first neighborhood of the first point cloud data point.

[0045] Then, in addition to using the radius method to construct the first neighborhood, the K nearest neighbor (KNN) method is also needed to obtain the second neighborhood. The specific operation is to start from the first point cloud data point and find its nearest K point cloud data points. The point cloud data points are sorted according to the distance from the first point cloud data point. The nearest K points constitute the second neighborhood, ensuring that the local characteristics of the point cloud data are fully considered from different angles.

[0046] Then, after obtaining the first neighborhood and the second neighborhood, the two neighborhoods need to be weighted and integrated. Using the coefficient of variation method, the importance of the points in each neighborhood is determined by calculating the coefficient of variation of each point in the two neighborhoods, and a weighted calculation is performed. The weighted results are combined to form a comprehensive neighborhood set. By considering the different characteristics of the two neighborhoods, the representativeness and reliability of the neighborhood set are improved.

[0047] Next, the neighborhood set is mathematically processed. First, the covariance matrix of the neighborhood set is calculated. The covariance matrix describes the distribution characteristics of the points in the neighborhood. Then, the covariance matrix is ​​eigenvalue decomposed to obtain eigenvalues ​​and eigenvectors. The eigenvector represents the main direction of the points in the neighborhood, and the eigenvector corresponding to the minimum eigenvalue is the first normal vector required. The geometric information in the neighborhood is extracted through mathematical operations, which provides the necessary data support for the subsequent feature point recognition.

[0048] By obtaining the first normal vector, it is ensured that the neighborhood of each point in the point cloud data is fully calculated and analyzed, so as to obtain an accurate normal vector, which provides a solid foundation for the subsequent point cloud feature point extraction and ensures the accuracy and reliability of data processing.

[0049] Furthermore, the present application also includes: According to the point cloud feature points, a nearest neighbor match is defined in the BIM model feature points, and the nearest neighbor match is performed to obtain the nearest matching point; the descriptor distance between the point cloud feature point and the BIM model feature point in the nearest matching point is calculated to obtain a similarity result, and a preliminary matching result is obtained according to the similarity result.

[0050] Specifically, the feature points extracted from the point cloud data are used as the source point set, and the feature points extracted from the BIM model are used as the target point set. In order to find the nearest neighbor point of each point cloud feature point in the BIM model feature point, it is necessary to use a nearest neighbor search algorithm, such as K nearest neighbor (KNN) or a KD tree-based search algorithm. Through the search algorithm, each point cloud feature point is traversed to find the BIM model feature point closest to it and marked as the nearest matching point, which can ensure that each point cloud feature point finds a closest BIM model feature point as a matching object.

[0051] Then, after finding the nearest neighbor matching point, the next step is to evaluate the similarity of the nearest neighbor matching point. The similarity evaluation is achieved by calculating the descriptor distance between the point cloud feature point and the BIM model feature point. Commonly used descriptors include FPFH and SHOT, which encode the local geometric information of the feature point into a feature vector. By calculating the Euclidean distance between the descriptor vectors of two feature points, the similarity between the two feature points can be quantified. The smaller the distance, the higher the similarity, indicating that the two feature points are more similar in geometry.

[0052] Then, all the nearest matching points are sorted and screened according to the calculated descriptor distance. A similarity threshold is set, and only matching point pairs with a descriptor distance less than the similarity threshold are considered valid preliminary matching results, ensuring that only feature point pairs with highly similar geometric shapes are retained, thereby improving the accuracy and reliability of matching. By analyzing the similarity results, those matching pairs that are nearest neighbors but inconsistent in geometry can be effectively filtered out, and finally a set of high-quality preliminary matching results can be obtained.

[0053] The matching points are determined through nearest neighbor search, the similarity is calculated through descriptor distance, and the preliminary matching results are screened based on the similarity, providing a solid foundation for subsequent feature point alignment and mapping.

[0054] Furthermore, the present application also includes: Partial matching pairs are sampled and selected from the preliminary matching results, a rigid transformation matrix is ​​calculated for the partial matching pairs, and the transformation matrix is ​​applied to the point cloud feature points to calculate the transformation distance between the transformed point cloud feature points and the BIM model feature points; if the transformation distance is less than a preset transformation distance threshold, the number of partial matching pairs that conform to the transformation matrix is ​​obtained; the sampling and transformation matrix are repeatedly iterated, and the transformation matrix with the largest number of partial matching pairs that conform to the transformation matrix is ​​selected as the optimal transformation matrix; the preliminary matching results are traversed and calculated according to the optimal transformation matrix to obtain a minimum distance error result, and an optimized matching result corresponding to the minimum distance error result is generated.

[0055] Specifically, a portion of matching pairs is randomly selected from the preliminary matching results, and some matching pairs contain point cloud feature points and their corresponding BIM model feature points. Then, the rigid transformation matrix is ​​calculated using some matching pairs, which includes rotation and translation operations, aiming to align the point cloud feature points to the BIM model feature points. After calculating the transformation matrix, it is applied to all point cloud feature points to obtain the transformed point cloud feature point positions. Next, the distance between the transformed point cloud feature points and the corresponding BIM model feature points is calculated, reflecting the alignment effect of the transformation matrix.

[0056] Then, if the transformation distance is less than the preset transformation distance threshold, the number of partial matching pairs that conform to the transformation matrix is ​​obtained. In order to evaluate the effect of the rigid transformation matrix, a preset transformation distance threshold needs to be set. Calculate the distance between the transformed point cloud feature points and the BIM model feature points, and compare the transformed distance with the preset threshold. If the transformation distance is less than the threshold, the matching pair is considered to conform to the transformation matrix. Count the number of all matching pairs that conform to the transformation matrix as a criterion for evaluating the quality of the transformation matrix. The more matching pairs that conform, the better the alignment effect of the transformation matrix.

[0057] Next, the sampling and transformation matrix are iterated repeatedly, and the transformation matrix with the largest number of partial matching pairs that conform to the transformation matrix is ​​selected as the optimal transformation matrix. In order to find the optimal rigid transformation matrix, multiple iterative sampling and transformation matrix calculations are required. In each iteration, a part of the matching pairs is randomly selected from the preliminary matching results, a new transformation matrix is ​​calculated, and its alignment effect is evaluated. Through multiple iterations, multiple transformation matrices and their corresponding number of matching pairs can be obtained. Finally, the transformation matrix with the largest number of matching pairs is selected as the optimal transformation matrix, ensuring that the best transformation matrix that can maximize the alignment of point cloud feature points and BIM model feature points is selected.

[0058] Then, once the optimal transformation matrix is ​​found, the distance error traversal calculation is performed on the preliminary matching results. Specifically, the optimal transformation matrix is ​​applied to all point cloud feature points in the preliminary matching results to calculate the distance error between the corresponding BIM model feature points. By traversing and calculating the distance error of all matching pairs, the set of matching pairs with the smallest error is found to form the final optimized matching result, ensuring that the alignment error between the point cloud feature points and the BIM model feature points is minimized, thereby improving the alignment accuracy and reliability.

[0059] By sampling and selecting some matching pairs from the preliminary matching results to calculate the rigid transformation matrix, evaluating its alignment effect, performing multiple iterative sampling and selecting the optimal transformation matrix, and finally using the optimal transformation matrix to perform distance error traversal calculation on the preliminary matching results, an optimized matching result corresponding to the minimum distance error is generated. Through continuous optimization and verification, high-precision alignment of point cloud feature points and BIM model feature points is ensured, providing a solid foundation for subsequent analysis and application.

[0060] Furthermore, the present application also includes: The centroids are calculated based on the feature point matching pairs in the optimized matching results, and the centroids are removed, the covariance matrix and singular value decomposition are constructed, and a preliminary rigid transformation is performed to obtain a rough alignment result; a spatial index is established for the rough alignment result, and the nearest neighbor point is found and the Euclidean distance of the nearest neighbor point is calculated to obtain an alignment error; the rigid transformation matrix is ​​iterated based on the alignment error to obtain a fine alignment result.

[0061] Specifically, all feature point matching pairs are extracted from the optimized matching results, including point cloud feature points and BIM model feature points. Then, the centroid, that is, the average position, of the point cloud feature points and BIM model feature points is calculated. Next, the centroid coordinates are subtracted from the coordinates of all feature points to perform a centroid removal operation to eliminate the effect of translation. After that, the covariance matrix is ​​constructed using the feature points after centroid removal, and the rotation matrix is ​​obtained through singular value decomposition (SVD). Finally, the rotation matrix and translation vector are combined to calculate the preliminary rigid transformation matrix, which is applied to the point cloud feature points to obtain a rough alignment result.

[0062] Then, a spatial index is established for the point cloud feature points in the rough alignment result to speed up the nearest neighbor search process. Common spatial indexing methods include KD trees or other efficient data structures. Then, for each transformed point cloud feature point, find its nearest neighbor point in the BIM model feature point. The nearest neighbor point refers to the point with the smallest distance in three-dimensional space. The Euclidean distance between each point cloud feature point and its corresponding nearest neighbor BIM model feature point is calculated as the alignment error. The alignment error reflects the distance deviation between the point cloud feature point and the BIM model feature point.

[0063] Then, in each iteration, the covariance matrix of the feature point set after removing the centroid is recalculated and singular value decomposition is performed to obtain a new rotation matrix and translation vector. Then, the new transformation matrix is ​​applied to transform the point cloud feature points and the alignment error is calculated again. Through continuous iteration, the total alignment error of all feature point pairs is gradually minimized until the error converges to the preset threshold or the maximum number of iterations is reached. Finally, an accurate fine alignment result is obtained, ensuring high-precision alignment between the point cloud feature points and the BIM model feature points.

[0064] The rough alignment result is obtained by calculating the centroid, removing the centroid, constructing the covariance matrix and singular value decomposition of the feature point matching pairs in the optimized matching results; then the spatial index is established, the nearest neighbor point is found and the Euclidean distance is calculated to obtain the alignment error; finally, the rigid transformation matrix is ​​iteratively updated according to the alignment error to obtain the fine alignment result, ensuring the high-precision alignment of the point cloud feature points and the BIM model feature points, providing a reliable foundation for further data processing and application.

[0065] Furthermore, the present application also includes: The residual is calculated according to the fine alignment result, and the residual is statistically analyzed to obtain an error analysis result; the fine alignment result is visually tested, and manual correction is performed to obtain a manual correction result; the correction result is obtained according to the error analysis result and the manual correction result.

[0066] Specifically, all matching pairs are extracted from the fine alignment results, including point cloud feature points and BIM model feature points. For each pair of matching points, the residual between the matching pairs is calculated, that is, the distance error between the point cloud feature point and the corresponding BIM model feature point. Then, all residuals are statistically analyzed, and statistical indicators such as the average residual, maximum residual, minimum residual, and standard deviation are calculated, which reflect the accuracy and error distribution of the alignment between the point cloud data and the BIM model. Through error analysis, the main sources of error in the alignment process can be identified and the overall alignment effect can be evaluated.

[0067] Then, the fine alignment results are displayed through visualization tools, including the alignment of point cloud feature points and BIM model feature points. Visual testing can help visually check the alignment effect and find possible alignment errors or anomalies. By viewing the visualization results, you can find subtle alignment errors or unreasonable matching points. After the problem is found, manual correction is performed to manually adjust the position of the feature points to align them more accurately. The manual correction results record the detailed information of each manual adjustment, including the adjusted feature points and the error changes after the adjustment. Through manual correction, the alignment accuracy can be further improved and the deficiencies of the automatic alignment algorithm can be supplemented.

[0068] Then, the overall alignment effect is comprehensively evaluated by combining the error analysis results and manual correction results. The error analysis results provide an accuracy evaluation of the automatic alignment algorithm and reveal the main error sources and distribution. The manual correction results provide the parts that need to be manually adjusted during the alignment process and record the detailed information and effects of each adjustment. By combining the two results, the final correction result can be obtained, including the combined effects of automatic alignment and manual correction, ensuring high-precision alignment of point cloud feature points with BIM model feature points. The correction results not only provide the final error evaluation of the alignment, but also record every step of the entire alignment process, providing detailed data support for subsequent quality control and further applications.

[0069] The residuals are calculated from the fine alignment results and statistically analyzed to obtain the error analysis results. The fine alignment results are then visually tested and manually corrected to obtain the manual correction results. Finally, the error analysis results and manual correction results are combined to obtain a comprehensive correction result, ensuring high-precision alignment of the feature points of the point cloud and the feature points of the BIM model, and providing a reliable data basis for subsequent applications and quality control.

[0070] In summary, the prefabricated building quality monitoring system based on BIM and point cloud integration provided by this application has the following technical effects: The BIM model feature point acquisition module is used to acquire the target BIM model according to the target prefabricated building, extract geometric features from the target BIM model, and obtain BIM model feature points; the point cloud feature point acquisition module is used to acquire the target point cloud data according to the target prefabricated building, extract point cloud features from the target point cloud data, and obtain point cloud feature points; the preliminary matching result acquisition module is used to perform descriptor matching based on the BIM model feature points and the point cloud feature points to obtain matching pairs, and perform matching pair screening to obtain preliminary matching results; the optimized matching result acquisition module is used to evaluate and transform the preliminary matching results, and perform error calculation to obtain The module is used to perform preliminary transformation calculation on the optimized matching result to obtain a rough alignment result, perform nearest neighbor search and minimize the error on the rough alignment result to obtain a fine alignment result; the module is used to perform error analysis and manual correction on the fine alignment result to obtain the correction result, monitor and generate a quality monitoring report based on the integrated whole process, and perform quality traceability through the quality monitoring report to generate quality improvement suggestions, thereby achieving the technical goals of optimizing feature point extraction, improving the robustness of the matching algorithm and strengthening error correction, and achieving the technical effect of significantly improving the alignment and mapping accuracy of point cloud data and BIM models.

[0071] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0072] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. The prefabricated building quality monitoring system based on BIM and point cloud integration is characterized by: include: A BIM model feature point acquisition module is used to acquire a target BIM model according to a target prefabricated building, extract geometric features from the target BIM model, and obtain BIM model feature points; A point cloud feature point acquisition module is used to acquire target point cloud data according to the target prefabricated building, extract point cloud features from the target point cloud data, and obtain point cloud feature points; A preliminary matching result obtaining module is used to perform descriptor matching based on the BIM model feature points and the point cloud feature points to obtain matching pairs, and to screen the matching pairs to obtain preliminary matching results; An optimized matching result obtaining module is used to evaluate and transform the preliminary matching result, and perform error calculation to obtain an optimized matching result; A fine alignment result obtaining module is used to perform preliminary transformation calculation on the optimized matching result to obtain a rough alignment result, perform nearest neighbor search and minimize error on the rough alignment result to obtain a fine alignment result; The quality improvement suggestion generation module is used to perform error analysis and manual correction on the fine alignment results, obtain correction results, monitor and generate a quality monitoring report based on the integrated full process, and perform quality traceability through the quality monitoring report to generate quality improvement suggestions.

2. The prefabricated building quality monitoring system based on BIM and point cloud integration according to claim 1 is characterized in that: The BIM model feature point acquisition module includes: Corner point extraction, edge extraction and plane intersection line extraction are performed according to the target BIM model to obtain geometric feature points; The geometric feature points are described according to feature descriptors, and neighborhood calculation of the geometric descriptors is performed to obtain BIM neighborhood feature distribution, the geometric feature points are marked in the target BIM model, and the BIM model feature points are obtained in combination with the BIM neighborhood feature distribution.

3. The prefabricated building quality monitoring system based on BIM and point cloud integration according to claim 1 is characterized in that: The point cloud feature point acquisition module includes: Extracting a first point cloud data point according to the target point cloud data; Performing neighborhood construction and normal vector calculation on the first point cloud data point to obtain a first normal vector; Calculating the curvature according to the rate of change of the first normal vector to obtain a first curvature value; Calculate the distance change between the first point cloud data point and the first neighborhood, and if the change rate exceeds a preset change rate threshold, obtain an edge point judgment result; Screening the first curvature value according to the curvature threshold, and obtaining the first feature point in combination with the edge point judgment result; A neighborhood calculation of the point cloud descriptor is performed according to the first feature point to obtain a first point cloud neighborhood feature distribution, the point cloud feature point is marked in the target point cloud data, and a first point cloud feature point is obtained in combination with the first point cloud neighborhood feature distribution, and the first point cloud feature point is added to the point cloud feature point.

4. The prefabricated building quality monitoring system based on BIM and point cloud integration according to claim 3 is characterized in that: The point cloud feature point acquisition module further includes: Acquire all point cloud data points within a preset point cloud radius of the first point cloud data point to generate a first neighborhood of the first point cloud data point; Acquire K point cloud data points closest to the first point cloud data point to generate a second neighborhood of the first point cloud data point; Performing weighted calculation on the first neighborhood and the second neighborhood based on a coefficient of variation method to obtain a neighborhood of the first point cloud data point, performing a neighborhood search to obtain a neighborhood set of the first point cloud data point; The covariance matrix of the neighborhood set is calculated, and eigenvalue decomposition is performed to obtain eigenvalues ​​and eigenvectors, thereby obtaining the first normal vector.

5. The prefabricated building quality monitoring system based on BIM and point cloud integration according to claim 1 is characterized in that: The preliminary matching result obtaining module includes: Defining nearest neighbor matching in the BIM model feature points according to the point cloud feature points, and performing nearest neighbor matching to obtain the nearest matching point; The descriptor distance between the point cloud feature point and the BIM model feature point in the nearest matching point is calculated to obtain a similarity result, and a preliminary matching result is obtained according to the similarity result.

6. The prefabricated building quality monitoring system based on BIM and point cloud integration according to claim 1 is characterized in that: The optimized matching result obtaining module includes: Sampling and selecting some matching pairs from the preliminary matching results, calculating a rigid transformation matrix for the some matching pairs, applying the transformation matrix to the point cloud feature points, and calculating the transformation distance between the transformed point cloud feature points and the BIM model feature points; If the transformation distance is less than a preset transformation distance threshold, the number of partial matching pairs that conform to the transformation matrix is ​​obtained; Repeatedly iterate sampling and transforming the matrix, and select the transformation matrix with the largest number of partial matching pairs that conform to the transformation matrix as the optimal transformation matrix; The distance error traversal calculation is performed on the preliminary matching result according to the optimal transformation matrix to obtain the minimum distance error result, and the optimized matching result corresponding to the minimum distance error result is generated.

7. The prefabricated building quality monitoring system based on BIM and point cloud integration according to claim 1 is characterized in that: The fine alignment result obtaining module comprises: Calculating the centroid of the feature point matching pairs in the optimized matching result, removing the centroid, constructing the covariance matrix and singular value decomposition, and performing preliminary rigid transformation to obtain a rough alignment result; Establishing a spatial index for the rough alignment result, finding the nearest neighbor point and calculating the Euclidean distance of the nearest neighbor point to obtain the alignment error; The rigid transformation matrix is ​​iterated according to the alignment error to obtain a fine alignment result.

8. The prefabricated building quality monitoring system based on BIM and point cloud integration according to claim 1 is characterized in that: The quality improvement suggestion generation module includes: Calculating residuals according to the fine alignment results, and performing statistical analysis on the residuals to obtain error analysis results; Performing a visual test on the fine alignment result, and performing manual correction to obtain a manual correction result; The correction result is obtained according to the error analysis result and the manual correction result.

Images