A prefabricated reinforced concrete component model generation method, system and device

Abstract

The application provides an assembled reinforced concrete component model generation method, system and equipment, comprising: obtaining three-dimensional point cloud data of an assembled reinforced concrete component; clustering and segmenting the three-dimensional point cloud data according to geometric features extracted from the three-dimensional point cloud data to obtain first point clouds corresponding to concrete and second point clouds corresponding to steel segments; performing multi-plane fitting on the first point clouds to obtain plane equations corresponding to the concrete; performing spatial multi-segment line fitting on the second point clouds to obtain linear parameters corresponding to the steel segments, and making the fitting straight lines corresponding to adjacent steel segments intersect under a preset constraint condition to obtain bending points; constructing a closed concrete entity model based on the plane equations, and constructing a steel entity model based on the linear parameters and the bending points; and generating a finite element model of the assembled reinforced concrete component according to the steel entity model and the concrete entity model. The application effectively improves modeling efficiency, reduces manual errors and costs.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and specifically to a method, system, and equipment for generating models of prefabricated reinforced concrete components. Background Technology

[0002] During the production and construction of RC (Reinforced Concrete) components (such as precast columns, precast beams, precast composite beams, and precast slabs), point cloud data is often obtained through 3D laser scanning. Converting this point cloud data into finite element (FE) models that can be used for structural analysis is a key step in automating the point cloud-to-analysis process for precast buildings.

[0003] In the field of point cloud to finite element modeling, reverse modeling and damage assessment of building structures are mainly achieved through 3D computer vision methods. For example, a general clustering algorithm is used to perform coarse-grained segmentation of the component point cloud, and macroscopic geometric parameters are fitted by methods such as slicing or Gaussian sphere spatial mapping to generate a finite element surface model for damage assessment. Other methods combine geometric features (such as linearity and flatness) with machine learning to identify the second point cloud and the concrete plane. In recent years, zero-shot segmentation methods based on large language models and registration methods that introduce design cross-sectional contours as geometric priors have also emerged, attempting to achieve component identification and reconstruction without relying on labeled data or sparse point clouds.

[0004] However, the above-mentioned technologies still have the following shortcomings: First, existing segmentation strategies lack fine-grained differentiation capabilities for materials with significantly different geometric features, such as steel reinforcement and concrete, making it difficult to simultaneously and accurately reconstruct the concrete surface and steel reinforcement skeleton; Second, supervised methods rely on a large amount of labeled data, while high-quality labeled point clouds are extremely scarce in the field of prefabricated RC components, limiting the generalization of the methods; Third, a complete automated pipeline from point cloud recognition to 3D reconstruction to model generation has not been established, and manual intervention or manual format conversion in professional software is still required in the process. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, system and equipment for generating prefabricated reinforced concrete component models to solve the technical problem that the existing technology cannot achieve the fully automated generation of finite element models of prefabricated reinforced concrete components from three-dimensional point cloud data.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a method for generating a model of a prefabricated reinforced concrete component, comprising: Acquire 3D point cloud data of prefabricated reinforced concrete components; Based on the geometric features extracted from the three-dimensional point cloud data, the three-dimensional point cloud data is clustered and segmented to obtain the first point cloud corresponding to the concrete and the second point cloud corresponding to the steel bar segment. Multiplane fitting is performed on the first point cloud to obtain the plane equations corresponding to each concrete. Spatial polyline fitting is performed on the second point cloud to obtain the linear parameters corresponding to each steel bar segment, and the fitted lines corresponding to adjacent steel bar segments intersect under preset constraints to obtain bending points. A closed concrete solid model is constructed based on the plane equation, and a steel reinforcement solid model is constructed based on the straight line parameters and the bending points. Based on the steel reinforcement solid model and the concrete solid model, a finite element model of the prefabricated reinforced concrete component is generated.

[0007] In one possible implementation, the step of clustering and segmenting the three-dimensional point cloud data based on geometric features extracted from the three-dimensional point cloud data to obtain a first point cloud corresponding to concrete and a second point cloud corresponding to the steel reinforcement segment includes: Obtain the set of neighborhood points within the neighborhood range of each point in the three-dimensional point cloud data; Calculate the neighborhood centroid of each point based on the neighborhood point set, and construct the covariance matrix of each point based on the neighborhood centroid and the neighborhood point set; The covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvalues; Calculate the linearity, flatness, sphericity, total variance, anisotropy, feature entropy, and curvature change of each point in the three-dimensional point cloud data based on the feature values. After normalizing the spatial coordinates of each point, they are concatenated with the linearity, flatness, sphericity, total variance, anisotropy, feature entropy, and curvature change to form a combined feature vector; Based on the combined feature vector, the points in the three-dimensional point cloud data are clustered to divide the point cloud into high flatness clusters and high linearity clusters, thereby obtaining the first point cloud and the candidate point cloud corresponding to the steel bar segment. The candidate point cloud is filtered using the plane equation obtained by multi-plane fitting. Points in the candidate point cloud whose distance to the nearest concrete plane is less than a preset threshold are removed from the candidate point cloud to obtain the second point cloud.

[0008] In one possible implementation, the step of performing multi-plane fitting on the first point cloud to obtain the plane equations corresponding to each concrete element includes: Three non-collinear points are randomly sampled from the first point cloud, and the candidate plane equation is calculated based on the sampled three non-collinear points; Count the number of target points in the first point cloud whose distance to the candidate plane equation is less than a preset threshold; The process of iteratively performing the random sampling, calculating the candidate plane equation, and counting the number of target points is used, and the candidate plane equation with the largest number of points is retained as the final plane equation. Remove points whose distance to the final plane equation is less than a preset threshold from the first point cloud. Repeat the above process for the remaining point cloud until the preset maximum number of fitted planes is reached or the number of remaining points in the first point cloud is less than a preset threshold, and obtain the plane equation corresponding to each concrete.

[0009] In one possible implementation, the step of performing spatial polyline fitting on the second point cloud to obtain the linear parameters corresponding to each steel bar segment includes: A density clustering algorithm is used to cluster the second point cloud with a pre-set neighborhood radius to obtain multiple single second point cloud clusters; the neighborhood radius is set according to the diameter of the reinforcing bar. Spatial polyline fitting is performed on each of the single second point cloud clusters to obtain the linear parameters corresponding to each steel bar segment.

[0010] In one possible implementation, the step of causing the fitted straight lines corresponding to adjacent steel bar segments to intersect under preset constraints to obtain a bend point includes: Obtain the direction vector of the first fitted line corresponding to the steel bar segment in the first round of fitting, as well as the known points of the first fitted line; Obtain the direction vector of the straight line to be fitted corresponding to the steel bar segment adjacent to the steel bar segment to be fitted in the first round of fitting; The normal vector is determined based on the direction vector of the first fitted line and the direction vector of the line to be fitted; The plane with the normal vector as its normal direction and passing through the known point is used as the constraint plane; The random sampling points of the line to be fitted are projected onto the constraint plane to obtain the projection points; Using the projection point as the starting point and the direction vector of the line to be fitted as the direction, a corrected line to be fitted is generated. The bending point is obtained from the intersection of the corrected line to be fitted and the first fitted line.

[0011] In one possible implementation, constructing a closed concrete solid model based on the plane equation includes: Determine the plane corresponding to each concrete according to the plane equation corresponding to each concrete, and construct an axisymmetric bounding box for each plane respectively. An intersection test is performed on each of the axisymmetric bounding boxes. Based on the intersection test results, the intersecting plane pairs are determined, and the intersection line between the intersecting plane pairs is calculated. The endpoints of each intersection line are determined according to a preset threshold sequence, and the boundaries of each plane are formed by the intersection lines and their endpoints. The concrete solid model is formed based on the boundaries of each plane.

[0012] In one possible implementation, constructing the solid model of the reinforcing steel bars based on the straight line parameters and the bend points includes: A oriented bounding box is constructed using the direction vector in the linear parameters corresponding to each steel bar segment as the main axis direction. The central axis of the oriented bounding box is determined based on the position information in the straight line parameters corresponding to each steel bar segment and the bending point. Calculate the cross-sectional radius of the directed bounding box in the direction perpendicular to the principal axis; Using the central axis as the center line of the reinforcing bar and the cross-sectional radius as the radius of the reinforcing bar, a solid model of the reinforcing bar corresponding to each reinforcing bar segment is generated.

[0013] In one possible implementation, generating the finite element model of the prefabricated reinforced concrete member based on the steel reinforcement solid model and the concrete solid model includes: The steel reinforcement solid model is embedded into the concrete solid model to form the finite element model; the steel reinforcement solid model and the concrete solid model interact with each other through finite element contact constraints or embedded constraints; the volume mesh of the finite element model is a tetrahedral volume mesh.

[0014] Secondly, the present invention also provides a prefabricated reinforced concrete component model generation system, comprising: The point cloud acquisition module is used to acquire three-dimensional point cloud data of prefabricated reinforced concrete components. The semantic segmentation module is used to perform clustering and segmentation on the three-dimensional point cloud data based on the geometric features extracted from the three-dimensional point cloud data to obtain the first point cloud corresponding to the concrete and the second point cloud corresponding to the steel bar segment. The fitting module is used to perform multi-plane fitting on the first point cloud to obtain the plane equations corresponding to each concrete; to perform spatial polyline fitting on the second point cloud to obtain the linear parameters corresponding to each steel bar segment, and to make the fitting lines corresponding to adjacent steel bar segments intersect under preset constraints to obtain bending points. The model generation module is used to construct a closed concrete solid model based on the plane equation, and to construct a steel reinforcement solid model based on the straight line parameters and the bending points; and to generate a finite element model of the prefabricated reinforced concrete component based on the steel reinforcement solid model and the concrete solid model.

[0015] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the prefabricated reinforced concrete component model generation method described in any of the above implementations.

[0016] The beneficial effects of this invention are as follows: Firstly, by integrating geometric feature-driven point cloud clustering and segmentation with a multi-level fitting optimization strategy, this invention can efficiently separate concrete and steel reinforcement components from the chaotic original point cloud and accurately restore the key geometric parameters of each component. This significantly improves the automation level of reverse modeling and the utilization rate of initial data, reducing the risk of error accumulation caused by manual intervention. Furthermore, by applying intersection constraints to the fitted straight lines of adjacent steel reinforcement segments to obtain bending points, and combining this with the concrete plane equation to construct a mutually matching solid model, a geometrically consistent and topologically correct combined finite element model can be formed under a unified coordinate system. This effectively enhances the realism of the force transmission simulation at the steel reinforcement-concrete interface, improves the computational convergence and reliability of subsequent structural analysis, and avoids repetitive modeling work caused by geometric distortion or simplification. It achieves fully automated extraction of concrete plane parameters, effectively improving the efficiency of reverse modeling of prefabricated reinforced concrete components. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of an embodiment of the prefabricated reinforced concrete component model generation method provided by the present invention; Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of S102; Figure 3 For the present invention Figure 1 A schematic diagram of an embodiment of S103; Figure 4 For the present invention Figure 1 A schematic diagram of an embodiment of S104; Figure 5 For the present invention Figure 1 Another embodiment of the process flow diagram of S104; Figure 6 For the present invention Figure 1 A schematic diagram of an embodiment of S105; Figure 7 For the present invention Figure 1 Another embodiment of the process diagram of S105; Figure 8 A schematic diagram of an embodiment of the prefabricated reinforced concrete component model generation system provided by the present invention; Figure 9 A schematic diagram of an embodiment of the electronic device provided by the present invention; Figure 10 A schematic flowchart of an embodiment of the prefabricated reinforced concrete component model generation method provided by the present invention; Figure 11 This is a schematic diagram of the clustering results of the point cloud of the prefabricated composite beam after improved K-Means unsupervised semantic segmentation in one embodiment of the present invention; Figure 12 This is a schematic diagram of the geometric reconstruction results after multi-plane fitting of precast composite beam concrete and RANSAC multi-segment fitting of steel reinforcement in one embodiment of the present invention; Figure 13 This is a schematic diagram illustrating the principle of the inner cylinder of the OBB in one embodiment of the present invention; Figure 14 This is a schematic diagram of the final finite element mesh model of the prefabricated composite beam in one embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] Before demonstrating the embodiments, the following terms will be explained.

[0024] Precast reinforced concrete components: These refer to "concrete structures assembled from precast concrete components using reliable connection methods." In simpler terms, they are prefabricated parts manufactured in a factory. Common types of precast reinforced concrete components include: precast columns, precast beams, precast composite beams, precast slabs, precast shear walls, and precast stairs.

[0025] FEM (Finite Element Model): A model established using the finite element analysis method. It is a combination of elements connected only at the nodes, transmitting forces only through the nodes, and constrained only at the nodes. It is the result of discretizing the mechanical model and is a digital model for numerical calculation.

[0026] This invention provides a method, system, and equipment for generating prefabricated reinforced concrete component models, which will be described below.

[0027] Figure 1 This is a schematic flowchart of an embodiment of the prefabricated reinforced concrete component model generation method provided by the present invention, as shown below. Figure 1 As shown, the method for generating prefabricated reinforced concrete component models includes: S101. Obtain the three-dimensional point cloud data of the prefabricated reinforced concrete component.

[0028] It should be noted that: by using non-contact sensing methods such as 3D laser scanning or structured light photogrammetry, dense spatial coordinate information of exposed steel bars on the surface and inside of prefabricated reinforced concrete components is collected to form an original 3D point cloud dataset.

[0029] S102. Based on the geometric features extracted from the three-dimensional point cloud data, the three-dimensional point cloud data is clustered and segmented to obtain the first point cloud corresponding to the concrete and the second point cloud corresponding to the steel bar segment.

[0030] It should be noted that after preprocessing the original point cloud, including denoising, downsampling, and normal vector calculation, local geometric features of each point (such as curvature, normal variation, local linearity, and flatness) are extracted. Based on these features, region growing or cluster-based segmentation algorithms are used to divide the point cloud data into two categories: one category corresponds to the point set of the concrete matrix region (first point cloud), and the other category corresponds to the point set of all steel reinforcement segments on the surface (second point cloud), thereby achieving effective separation of material components in geometric space.

[0031] S103. Perform multi-plane fitting on the first point cloud to obtain the plane equations corresponding to each concrete.

[0032] It should be noted that: for the first point cloud (concrete region) obtained from segmentation, the Random Sample Consensus (RANSAC) or least squares fitting algorithm is used to perform multi-plane fitting on each of the main outer surfaces of the component (including side, end, top, and bottom surfaces), and the mathematical equations (normal vector and intercept) of each plane are calculated one by one. During the fitting process, orthogonal or parallel constraints between adjacent planes need to be considered to improve geometric consistency and provide accurate boundary conditions for the subsequent construction of closed solids.

[0033] S104. Perform spatial polyline fitting on the second point cloud to obtain the straight line parameters corresponding to each steel bar segment, and make the fitted straight lines corresponding to adjacent steel bar segments intersect under preset constraint conditions to obtain bending points.

[0034] It should be noted that the second point cloud (reinforcing bar segment) is segmented according to connectivity or direction. For each continuous reinforcing bar segment, a three-dimensional spatial straight line or circular arc fitting is performed to obtain the axial direction and position parameters of each segment. Based on this, by introducing the intersection constraint of the fitting lines of adjacent reinforcing bar segments in physical space (such as least squares intersection under the condition of allowing small angular deviations), the spatial coordinates of each bending point are accurately calculated, thereby connecting the discrete reinforcing bar segments into a continuous spatial polygonal skeleton, which accurately reflects the bending direction and anchorage form of the reinforcing bar.

[0035] S105. Construct a closed concrete solid model based on the plane equation, and construct a steel reinforcement solid model based on the straight line parameters and the bending points.

[0036] It should be noted that: using all concrete plane equations, a non-self-intersecting, volume-closed concrete solid model (which can be represented using B-rep or CSG) is generated through plane trimming, Boolean operations, and shell closure. Simultaneously, based on the axial parameters and bending point coordinates of each rebar segment, a circular cross-section is assigned according to the actual rebar diameter, and a complete rebar solid model is formed by sweeping along the axis. Both types of models must maintain a consistent geometric coordinate system, and the rebar solid should be located within the corresponding concrete cover thickness of the concrete solid.

[0037] S106. Generate the finite element model of the prefabricated reinforced concrete component based on the steel reinforcement solid model and the concrete solid model.

[0038] It should be noted that the above concrete solid model and steel reinforcement solid model are spatially assembled and combined. According to the analysis requirements (such as structural static, dynamic or nonlinear analysis), the mapping mesh or adaptive tetrahedral / hexahedral mesh generation technology is used to generate finite element mesh elements for the concrete domain and the steel reinforcement domain respectively, and the interface constraints between the two are defined (such as embedded constraints or common node coupling). Finally, a refined finite element model of the complete prefabricated reinforced concrete component is output.

[0039] In summary, the prefabricated reinforced concrete component model generation method provided in this embodiment of the invention firstly, by integrating geometric feature-driven point cloud clustering and segmentation with a multi-level fitting optimization strategy, can efficiently separate concrete and steel reinforcement components from the chaotic original point cloud and accurately restore the key geometric parameters of each component. This significantly improves the automation level of reverse modeling and the utilization rate of initial data, reducing the risk of error accumulation caused by manual intervention. Furthermore, by applying intersection constraints to the fitted straight lines of adjacent steel reinforcement segments to obtain bending points, and combining them with the concrete plane equation to construct a mutually matching solid model, a geometrically consistent and topologically correct combined finite element model can be formed under a unified coordinate system. This effectively enhances the realism of the force transmission simulation at the steel reinforcement-concrete interface, improves the computational convergence and reliability of subsequent structural analysis, and avoids repetitive modeling work caused by geometric distortion or simplification. It also achieves fully automated extraction of concrete plane parameters, effectively improving the efficiency of reverse modeling of prefabricated reinforced concrete components.

[0040] In some embodiments of the present invention, such as Figure 2 As shown, step S102 includes: S201. Obtain the set of neighboring points within the neighborhood range of each point in the three-dimensional point cloud data.

[0041] It should be noted that for each point in the input point cloud, its search radius is used to determine the neighborhood range, and the set of neighboring points within that range is obtained. That is, for each point in the 3D point cloud data, all neighboring points whose distance from the point is less than or equal to the search radius are quickly retrieved using a spatial indexing structure such as KD-Tree, and the indices or coordinates of these points are recorded as the neighborhood point set.

[0042] S202. Calculate the neighborhood centroid of each point based on the neighborhood point set, and construct the covariance matrix of each point based on the neighborhood centroid and the neighborhood point set.

[0043] It should be noted that for each point in the input point cloud p i Any point in a point cloud pi The coordinates in three-dimensional space are p i = ( x i , y i , z i ), based on its neighborhood search radius r= Point set within 40mm N i The neighborhood centroid is calculated as follows: ; in, p i For the first point cloud in the input point cloud i One point, u i For the first i The centroids (three-dimensional coordinate vectors) of the neighborhood of each point. K For neighborhood point set N i The number of internal target points p j For the first in the neighborhood j The three-dimensional coordinate vector of a point N i For the first i Points p i The set of all neighborhood points within the search radius r.

[0044] The covariance matrix is ​​then constructed as follows: ; in, For the first i The neighborhood covariance matrix of n points, where T is the transpose.

[0045] S203. Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues.

[0046] It should be noted that performing eigenvalue decomposition on a 3×3 covariance matrix can transform the spatial distribution information of the neighborhood point set into numerical eigenvalues, including λ1, λ2, and λ3, and the eigenvalues ​​satisfy the following: λ1≥λ2≥λ3≥0.

[0047] S204. Calculate the linearity, flatness, sphericity, total variance, anisotropy, feature entropy, and curvature change of each point in the three-dimensional point cloud data based on the feature values.

[0048] It should be noted that the 7-dimensional geometric features are calculated based on eigenvalues. These 7-dimensional geometric features include linearity, flatness, sphericity, total variance, anisotropy, eigenentiation, and curvature variation. The formula for calculating linearity is shown below: L λ,i =(λ1-λ2) / λ1; Among them, L λ,i For the first i The linearity corresponding to each point (reflecting the degree to which the point cloud has a linear structure; this value is larger for the second point cloud), λ1 is the first eigenvalue of the covariance matrix, and λ2 is the second eigenvalue of the covariance matrix.

[0049] The formula for calculating flatness is as follows: P λ,i =(λ2-λ3) / λ1; Among them, P λ,i For the first i The flatness of each point (reflecting the degree to which the point cloud has a planar structure; this value is larger for the first point cloud), and λ3 is the third eigenvalue of the covariance matrix.

[0050] The formula for calculating sphericity is as follows: S λ,i =λ3 / λ1; Among them, S λ,i For the first i The sphericity of each point (reflecting the degree to which the point cloud is spherically distributed).

[0051] The formula for calculating total variance is as follows: ; in, For the first i The total variance corresponding to each point (reflecting the overall dispersion of the neighborhood point cloud).

[0052] The formula for calculating anisotropy is shown below: A λ,i =(λ1-λ3) / λ1; in, A λ,i For the first i The anisotropy of each point (reflecting the non-uniformity of the point cloud along its direction).

[0053] The formula for calculating feature entropy is as follows: ; e m =λ m / (λ1+λ2+λ3) in, For the firsti The feature entropy corresponding to each point (reflecting the degree of disorder in the distribution of feature values). As an intermediate variable, the first m The proportion of each eigenvalue of the covariance matrix to the sum of all eigenvalues. m =1, 2, 3.

[0054] The formula for calculating curvature change is shown below: C λ,i =λ3 / (λ1+λ2+λ3) in, C λ,i For the first i The curvature change at each point (reflecting the degree of curvature of the local surface).

[0055] S205. After normalizing the spatial coordinates of each point, concatenate them with the linearity, flatness, sphericity, total variance, anisotropy, feature entropy, and curvature change to form a combined feature vector.

[0056] It should be noted that: Computational point cloud center The spatial coordinates are normalized to the interval [1, 1] by calculating the maximum distance dmax from the center. The normalized spatial coordinates are then concatenated with the 7-dimensional geometric features to form a 10-dimensional combined feature vector, as shown below: ; in, For the first i A 10-dimensional combined feature vector of points, For the first i The spatial normalized coordinates of the points.

[0057] in, The average value of all points in the point cloud in the X, Y, and Z directions, linearity, and flatness are assigned weights w. L =5 and w p =5, with the remaining features having a weight of 1. By normalizing the spatial coordinates and concatenating them with geometric features, the clustering algorithm can simultaneously consider the spatial location and local geometry of points. Assigning larger weights to linearity and flatness can strengthen the influence of these two most discriminative features in clustering, improve the segmentation accuracy between the second and first point clouds, and avoid clustering results biased towards spatial location due to excessive differences in the magnitude of coordinate values ​​and geometric features.

[0058] S206. Cluster the points in the three-dimensional point cloud data according to the combined feature vector, divide the point cloud into high flatness clusters and high linearity clusters, and obtain the candidate point cloud corresponding to the first point cloud and the rebar segment.

[0059] It should be noted that the K-Means algorithm divides the data into k clusters through iterative optimization, maximizing the similarity between sample points within the same cluster and minimizing the similarity between different clusters. In this embodiment, k=2, corresponding to the first point cloud (high flatness cluster) and the candidate second point cloud (high linearity cluster), respectively. An improved distance metric is used during clustering, considering the weights of each feature dimension when calculating the Euclidean distance (linearity and flatness are weighted at 5, and other features at 1), thus guiding the clustering algorithm to focus more on the key geometric features that distinguish concrete from steel reinforcement. The segmentation results are as follows: Figure 2 As shown, points on the concrete surface and points on the reinforcing bars are accurately distinguished.

[0060] S207. The candidate point cloud is screened using the plane equation obtained by multi-plane fitting. Points in the candidate point cloud whose distance to the nearest concrete plane is less than a preset threshold are removed from the candidate point cloud to obtain the second point cloud.

[0061] It should be noted that the plane equations corresponding to each concrete surface are obtained through RANSAC multi-plane fitting. For each point in the candidate point cloud, the distance from that point to all concrete planes is calculated, and the nearest concrete plane and its distance value are found. If the distance is less than a preset threshold (e.g., 3mm), the point is determined to be a mis-segmented concrete boundary point or corner point and is removed from the candidate point cloud. After this round of screening, the remaining candidate point cloud is the pure second point cloud.

[0062] In this embodiment, by extracting the eigenvalues ​​of the covariance matrix of each point's neighborhood and calculating 7-dimensional geometric features (linearity, flatness, sphericity, total variance, anisotropy, eigenentiation, and curvature variation), the local geometric morphology around each point can be comprehensively quantified, providing sufficient mathematical basis for the accurate distinction between the first and second point clouds. Furthermore, by normalizing the spatial coordinates and concatenating them with the 7-dimensional geometric features to form a 10-dimensional combined feature vector, and assigning high weights to linearity and flatness, the KMeans clustering algorithm simultaneously considers the spatial location and geometric morphology of points during the segmentation process, significantly improving the accuracy and robustness of the segmentation boundary between concrete and reinforcing steel. Furthermore, the candidate second point cloud is purified by using the concrete plane obtained through multi-plane fitting. This effectively removes points that are easily missegmented at the concrete boundary or corner, avoiding interference from missegmented points on subsequent single rebar separation and geometric parameter extraction. The entire semantic segmentation process does not rely on manually labeled training samples. It can accurately distinguish between concrete and rebar based solely on the geometric features of the point cloud itself, effectively reducing reliance on labeled data and professional experience and improving applicability.

[0063] In some embodiments of the present invention, such as Figure 3 As shown, step S103 includes: S301. Randomly sample three non-collinear points from the first point cloud, and calculate the candidate plane equation based on the sampled three non-collinear points.

[0064] It should be noted that the core idea of ​​the RANSAC (Random Sample Consensus) algorithm is to estimate mathematical model parameters through random sampling and hypothesis testing. In this embodiment, three non-collinear points are randomly selected from the first point cloud. p 1. p 2. p 3. These three points determine a candidate plane. Let any point on the plane be... H Given that the coordinates are (x, y, z), the plane normal vector is n = (a, b, c), and the distance from the origin to the plane is d, the equation of the candidate plane can be expressed as: π k : n k · H = d k ; in, π k For the first k A concrete plane, n k For the first k The unit normal vector of a plane. H Let be the three-dimensional coordinate vector of any point on the plane. d k From the origin to the th k The distance between planes.

[0065] Composed of three non-collinear points p 1. p 2. p 3. The specific method for calculating the plane parameters is as follows: First, calculate the two edge vectors. v 1= p 2- p 1 and v 2= p 3- p 1, then through the cross product n = v 1× v 2. Obtain the direction of the normal vector, and then normalize it to obtain the unit normal vector. n Then substitute any known point to calculate the distance.

[0066] S302. Count the number of target points in the first point cloud whose distance to the candidate plane equation is less than a preset threshold.

[0067] It should be noted that: for each point in the first point cloud, the distance from that point to the candidate plane is calculated. If the distance is less than a preset threshold, the point is marked as a target point (i.e., an interior point); otherwise, it is marked as an exterior point. The number of all target points that satisfy the condition of a distance less than the preset threshold is counted.

[0068] S303. Iteratively execute the random sampling, the calculation of candidate plane equations, and the counting of the number of target points, and retain the candidate plane equation with the largest number as the final plane equation.

[0069] It should be noted that in each iteration, three non-collinear points are independently and randomly sampled to generate candidate planes, and the number of target points in each candidate plane is counted. After multiple iterations, the number of target points in all candidate planes is compared, and the candidate plane with the most target points is selected as the optimal plane for RANSAC fitting in this iteration. The number of iterations should be set to ensure that samples consisting entirely of target points are sampled with a high probability; this is usually calculated automatically based on the proportion of target points and the number of sampled points.

[0070] S304. Remove the points whose distance to the final plane equation is less than a preset threshold from the first point cloud. Repeat the above process for the remaining point cloud until the preset maximum number of fitted planes is reached or the number of remaining points in the first point cloud is less than a preset threshold, and obtain the plane equation corresponding to each concrete.

[0071] It should be noted that the target points corresponding to the retained optimal planes (i.e., points whose distance is less than a preset threshold) are removed from the first point cloud. Processes S301 to S303 are repeated on the remaining point cloud until the preset maximum number of fitted planes Nmax = 20 or the number of remaining points in the first point cloud falls below a preset threshold, thus obtaining the plane equations corresponding to each concrete surface. This iterative strategy allows the algorithm to sequentially extract the plane equations corresponding to all major concrete planes from the first point cloud. As the number of iterations increases, the remaining point cloud gradually decreases. The iteration terminates when the remaining point cloud is insufficient to fit a valid plane or the preset maximum number of fitted planes has been reached.

[0072] In this embodiment, by randomly sampling three non-collinear points to generate candidate planes and iteratively optimizing them, the accurate plane equations corresponding to each concrete plane can be robustly extracted from the first point cloud containing noise and outliers. Furthermore, a target point count mechanism is used to screen the optimal plane with the highest support, effectively avoiding fitting bias caused by local noise points and ensuring the accuracy of plane fitting. Further, through iteration, all main planes can be sequentially identified from a point cloud containing multiple planes. Simultaneously, by setting a maximum number of fitted planes, priority is given to extracting the main structural planes when the point cloud quality is poor, thereby improving adaptability and robustness under extreme point cloud quality conditions. The entire multi-plane fitting process does not require manual specification of the number of planes and initial positions, achieving fully automated extraction of concrete plane parameters and effectively improving the efficiency of reverse modeling of prefabricated reinforced concrete components.

[0073] In some embodiments of the present invention, such as Figure 4 As shown, step S104 includes: S401. A density clustering algorithm is used to cluster the second point cloud with a pre-set neighborhood radius to obtain multiple single second point cloud clusters; the neighborhood radius is set according to the diameter of the reinforcing bar.

[0074] It should be noted that the DBSCAN algorithm is a density-based spatial clustering algorithm. Its core idea is to determine the cluster structure by the density of the sample points. For each point in the second point cloud, the algorithm searches for the number of neighboring points within its neighborhood radius. If this number exceeds the minimum number of samples, the point is marked as a core point, and the algorithm expands outward from this core point, grouping densely connected points into the same cluster.

[0075] The neighborhood radius is set based on the following criteria: ε =1.2 d ; in, ε The spatial neighborhood radius for DBSCAN density clustering. d The diameter of the reinforcing bar.

[0076] According to the requirement in the "Code for Design of Concrete Structures" (GB50010-2010) that the spacing of reinforcing bars should be ≥ 1.5 times the diameter, the neighborhood radius is set. ε =1.2 d With a minimum sample size of minPts=10, the independent point cloud clusters of each individual steel bar are separated.

[0077] S402. Perform spatial polyline fitting on each of the single second point cloud clusters to obtain the linear parameters corresponding to each steel bar segment.

[0078] It should be noted that: assuming the first round of fitted straight line l The parametric equation for 1 is shown below: l 1: p = p 1 +τ 1· v 1; in, l 1 represents the first-round fitted line, which is the longest line segment obtained from the first round of RANSAC fitting. p Let be the three-dimensional coordinate vector of any point on the line. p 1 represents the coordinate vector of a known point on the straight line (usually taken as the centroid of the point cloud segment or a RANSAC sampling point). τ 1 is the first scalar free parameter in the parametric equation of the line. v 1 represents the straight line segment. l 1 is the unit vector in the direction of the vector.

[0079] The parametric equation for any straight line is shown below: l : p = p i +τ · v ; in, l Let be any straight line in space. p i For this straight line l The coordinates of any known point selected above, τ For scalar free parameters, v For this straight line l Directional unit vector (| v | = 1).

[0080] point q The formula for calculating the distance to a line is as follows: d ( q , l ) = | ( q-p 0) × v |; in, d ( q , l ) is a point q to the straight line l distance, p 0 represents the external point from which the distance is to be calculated. q The three-dimensional coordinate vector.

[0081] Distance less than t LThe points that are not found are assigned to the target point set of the line segment and removed from the point cloud. For the remaining point cloud, an intersection constraint mechanism is introduced to fit subsequent line segments. Let the fitted line... l The parameter of 1 is p 1 and v 1. For the current line segment to be fitted l 2. Make it satisfy the following conditions: n = v 1× v 2; in, n The normal vector used to construct the constraint plane. v 1 represents the first round of fitted straight line. l 1 is the unit vector in the direction of the vector. v 2 represents the line segment to be fitted later. l 2 is the unit vector in the direction of the vector.

[0082] by n The normal vector passes through the point p The constraint plane Π of 1 is represented as follows: Π: n ·( p - p 1) = 0; Among them, Π represents the origin of... n and p 1. Defined constraint plane (ensuring) l 2 intersects the plane, and the intersection point falls on the same plane. l 1 (above).

[0083] Random sampling points p s Projecting onto plane Π yields p ´ s The corrected line to be fitted that satisfies the intersection condition is shown below: l ´2: p = p ´ s + τ 2. v 2; in, l ´2 represents the subsequent fitted line segment after projection constraint correction. p For the corrected straight line l The three-dimensional coordinate vector of any point on 2. p ´ s These are the fixed anchor points after the constraint (i.e., the base points that the corrected straight line must pass through). τ 2 is the second scalar free parameter.

[0084] Repeat this process until the remaining point cloud does not meet the linearity threshold.t L (For example t L =0.7). The linearity threshold is an empirical value, which can be determined by statistically analyzing the linearity distribution of a large number of second point cloud samples. The linearity of a typical second point cloud is generally in the range of 0.8 to 0.95, while the linearity near the bend point or in noisy areas drops significantly to below 0.5 to 0.7. This can be addressed by setting... t L A value of 0.7 ensures good fitting quality while avoiding overfitting to low-quality residual point clouds. The termination condition "not meeting the linearity threshold" means that the average linearity of the remaining point cloud is lower than... This indicates that the residual point cloud no longer possesses sufficient linear characteristics to continue fitting the straight line segments. The output parameters are the straight line parameters (pj, vj), bending point coordinates, and rebar diameters for each rebar segment. The results of concrete plane fitting and rebar skeleton fitting are as follows: Figure 12 As shown.

[0085] In this embodiment, the DBSCAN density clustering algorithm, based on the rebar spacing specifications, sets the neighborhood radius, reliably separating densely arranged rebars into single second point cloud clusters, effectively avoiding clustering errors caused by excessively small rebar spacing or point cloud noise. Furthermore, by introducing an intersection constraint mechanism in the RANSAC polyline fitting, adjacent straight line segments are forced to intersect precisely at spatial bending points, effectively avoiding the bending point misalignment problem caused by independent fitting of each segment in traditional RANSAC, significantly improving the geometric accuracy and spatial coherence of the rebar skeleton reconstruction. Furthermore, by setting a linearity threshold as the fitting termination condition, overfitting of low-quality residual point clouds is avoided while ensuring fitting quality, enhancing robustness against point cloud noise and local missing values. Moreover, the entire rebar parameter extraction process requires no manual intervention, achieving fully automated extraction from the second point cloud to straight line parameters and bending point coordinates, effectively improving the efficiency of reverse modeling of prefabricated reinforced concrete components.

[0086] In some embodiments of the present invention, such as Figure 5 As shown, step S104 further includes: S501. Obtain the direction vector of the first fitted line corresponding to the steel bar segment in the first round of fitting, and the known points of the first fitted line.

[0087] It should be noted that for a single second point cloud cluster, the first round of RANSAC (Random Sample Consensus) fitting randomly samples two points from the point cloud to determine a candidate straight line. The number of target points whose distance to this line is less than a linearity threshold is counted. After multiple iterations, the straight line with the largest number of target points is retained as the longest straight line segment in the first round of fitting. The parameterized form of this straight line segment is: l 1: p = p 1 +τ 1· v 1; in, l 1 represents the first-round fitted line, which is the longest line segment obtained from the first round of RANSAC fitting. p Let be the three-dimensional coordinate vector of any point on the line. p 1 represents the coordinate vector of a known point on the straight line (usually taken as the centroid of the point cloud segment or a RANSAC sampling point). τ 1 is the first scalar free parameter in the parametric equation of the line. v 1 represents the straight line segment. l 1 is the unit vector in the direction of the vector.

[0088] S502. Obtain the direction vector of the straight line to be fitted corresponding to the steel bar segment adjacent to the steel bar segment to be fitted in the first round of fitting.

[0089] It should be noted that: for the remaining point cloud of the same rebar (i.e., the remaining point cloud that was not assigned to the target point after the first round of fitting), RANSAC line fitting is performed again to obtain the direction unit vector of the current line segment to be fitted. v 2. This direction vector v 2. The direction of the line segment to be fitted in space is estimated from the main direction of the current second point cloud cluster by the RANSAC algorithm. Adjacent refers to the next line segment that is spatially connected to the line segment fitted in the first round in the direction of the reinforcement extension.

[0090] S503. Determine the normal vector based on the direction vector of the first fitted line and the direction vector of the line to be fitted.

[0091] It should be noted that: based on the direction vector of the first fitted line v 1. Direction vector of the line to be fitted v The cross product of 2 determines the normal vector. n .because v 1 and v Both 2 are unit vectors, and their cross product is... n To be perpendicular to v 1 and v A vector of 2. When v 1 and v When 2 are not parallel, n A non-zero vector can be used as the normal direction of the subsequent constraint plane; when v 1 and v When two segments are parallel, it means that the directions of the two adjacent steel bars are the same. At this time, there are no bending points and no need to perform intersection constraint treatment.

[0092] S504. The plane with the normal vector as the normal direction and passing through the known point is used as the constraint plane.

[0093] It should be noted that: with n The normal vector passes through the point p The constraint plane Π of 1 is represented as follows: Π: n ·( p - p 1) = 0; Among them, Π represents the origin of... n and p 1. Defined constraint plane (ensuring) l 2 intersects the plane, and the intersection point falls on the same plane. l 1 (above).

[0094] By using the normal vector n Constructing a constraint plane with the normal direction and passing through a known point ensures that this plane simultaneously contains the first-round fitted line, thus making the corrected line to be fitted coplanar with the first-round fitted line in space. Furthermore, two non-parallel lines under the condition of coplanarity must intersect, which provides a geometric guarantee for subsequently forcing adjacent line segments to intersect precisely at the bending point.

[0095] S505. Project the random sampling points of the line to be fitted onto the constraint plane to obtain projection points.

[0096] It should be noted that the original RANSAC algorithm, when fitting the line segment to be fitted, randomly samples two points from the current point cloud to determine the direction of the candidate line. One of the sampled points is denoted as... p s Another point is used to determine the direction vector. v 2. To satisfy the intersection constraint, it is necessary to... p s Project orthogonally onto the constraint plane.

[0097] S506. Using the projection point as the starting point and the direction vector of the line to be fitted as the direction, generate the corrected line to be fitted.

[0098] It should be noted that: random sampling points p s Projecting onto plane Π yields p ´ s The corrected line to be fitted that satisfies the intersection condition is shown below: l ´2: p = p ´ s + τ 2. v 2; in, l ´2 represents the subsequent fitted line segment after projection constraint correction. p For the corrected straight line l The three-dimensional coordinate vector of any point on 2. p ´ s These are the fixed anchor points after the constraint (i.e., the base points that the corrected straight line must pass through). τ 2 is the second scalar free parameter.

[0099] With the original line to be fitted l Compared to 2, the corrected line to be fitted l ´2 direction vector v 2 remains unchanged (ensuring the fitting direction is consistent with the main direction of the point cloud), but its position changes from passing through the original sampling point. p s Transformed through the projection point p ´ s .because p ´ s Located on the constraint plane, and v 2. Simultaneously perpendicular to the normal vector n ,therefore l ´2 The entire straight line lies on the constraint plane.

[0100] S507. The bending point is obtained based on the intersection of the corrected line to be fitted and the first fitted line.

[0101] It should be noted that: due to l 1 and l Both ´2 are on the constraint plane and have different directions ( v 1 and v (2. Not parallel) The two lines must intersect at a point in space, which is the bend point between adjacent steel reinforcement segments. The coordinates of the intersection point are obtained by solving the parametric equations of the two lines simultaneously: p 1 +τ 1· v 1 =p ´ s + τ 2. v 2; Since the two lines are coplanar and not parallel, the coordinates of the intersection point, i.e. the coordinates of the bend point, can be obtained by solving the parametric equations mentioned above.

[0102] Repeat steps S501 to S507 until the remaining point cloud of the single second point cloud cluster no longer meets the linearity threshold, indicating that the residual point cloud no longer has sufficient linear features to continue fitting the straight line segment. At this point, terminate the iteration.

[0103] In this embodiment, an intersection constraint mechanism is introduced to force adjacent line segments to intersect precisely at spatial bends, effectively avoiding the bend misalignment problem caused by independent fitting of each segment in traditional RANSAC, and significantly improving the geometric accuracy and spatial coherence of the rebar skeleton reconstruction. Furthermore, by determining the normal vector and constructing a constraint plane through cross product operation, the intersection problem of adjacent line segments is transformed into a line intersection problem in the plane. Geometrically, this ensures that the corrected line to be fitted will inevitably intersect the first-round fitted line, eliminating the need for additional iterative optimization or approximation, and greatly improving the accuracy and stability of bend point calculation. Furthermore, by setting a linearity threshold as the fitting termination condition, overfitting of low-quality residual point clouds is avoided while ensuring fitting quality, enhancing the algorithm's robustness to point cloud noise and local missing data. Finally, this method achieves accurate polyline fitting for rebar skeletons of arbitrary spatial shapes, providing accurate geometric parameters for subsequent reconstruction of the rebar solid model.

[0104] In some embodiments of the present invention, such as Figure 6 As shown, step S105 includes: S601. Determine the plane corresponding to each concrete according to the plane equation corresponding to each concrete, and construct an axisymmetric bounding box for each plane.

[0105] It should be noted that: the plane equations corresponding to each concrete plane have been obtained through RANSAC multiplane fitting, and the corresponding plane in space can be determined from the plane equations. The AABB bounding box (i.e., axisymmetric bounding box) is constructed based on the actual spatial distribution of the point cloud of each plane, that is, the minimum and maximum coordinate values ​​of the point cloud of each plane are calculated in the X, Y, and Z axes respectively, forming an axisymmetric cuboid bounding box.

[0106] S602. Perform an intersection test on each of the axisymmetric bounding boxes, determine the intersecting plane pairs based on the intersection test results, and calculate the intersection line between the intersecting plane pairs.

[0107] It should be noted that the criterion for determining intersection is that the projection intervals of the two bounding boxes overlap in all three axes. This is the standard method for determining intersection in the AABB test. For any two planes i and j, their corresponding AABB bounding boxes are Bi and Bj, respectively. If the projection intervals of Bi and Bj overlap in the X, Y, and Z axes, then the two planes are considered to intersect; otherwise, they are considered not to intersect. For planes determined to intersect, the line of intersection between them is calculated.

[0108] S603. Determine the endpoints of each intersection line according to a preset threshold sequence, form the boundary of each plane by enclosing the intersection lines and their endpoints, and form the concrete solid model based on the boundary of each plane.

[0109] It should be noted that: the judgment conditions are gradually relaxed according to the preset threshold sequence to determine the endpoints of each intersection line. The boundaries of each plane are formed by the intersection lines and their endpoints, and the boundaries of each plane together constitute a closed concrete solid model. The progressively expanding threshold list is represented as Elist=[0, 0.01, 0.02, 0.03, 0.04, 0.05m].

[0110] For each intersection line, the program first searches for the nearest other intersection line to its endpoint using the current threshold (starting from 0). If the distance between the endpoints of the two intersection lines is less than the current threshold, the two intersection lines are connected at that endpoint. If a connection cannot be completed under the current threshold, the threshold is increased to the next threshold in the sequence, and the program continues to try. The threshold table is used to allow the program to progressively extend the intersection lines, ensuring that a line intersects with its nearest other line. Otherwise, a line might intersect with a more distant line, resulting in an incorrect intersection point. This is a robustness guarantee strategy.

[0111] Once the endpoints of the intersection lines are determined, the boundaries of each plane are formed by enclosing the intersection lines and their endpoints. The meaning of surface-line-point topological reasoning is to start from the concrete plane ("surface"), determine which pairs of planes have intersection lines ("lines") through intersection tests, then determine the boundary corner points from the intersection points ("points") between the intersection lines, ultimately forming a closed concrete body boundary. The boundaries of each plane together constitute a closed concrete solid model.

[0112] In this embodiment, by constructing an axisymmetric bounding box for the point cloud corresponding to each concrete plane and performing an intersection test, the theoretically infinitely large plane is constrained into a finite-sized planar patch, effectively avoiding unnecessary intersection line calculations and significantly reducing computational overhead. Furthermore, by progressively relaxing the judgment conditions according to a preset threshold sequence to determine the intersection line endpoints, the problem of inaccurate coincidence of intersection line endpoints caused by point cloud noise and fitting errors is effectively solved, enhancing robustness and adaptability under real engineering scanning conditions. Furthermore, the surface-line-point topological reasoning method constructs intersection lines and intersection points layer by layer from the concrete plane, forming a complete geometric reasoning chain from surface to line to point, ensuring the accurate closure of the concrete solid model boundary and providing a precise geometric carrier for the subsequent embedding of the reinforcing steel solid model.

[0113] In some embodiments of the present invention, such as Figure 7 As shown, step S105 further includes: S701. Construct a oriented bounding box with the direction vector in the linear parameters corresponding to each steel bar segment as the main axis direction.

[0114] It should be noted that for each segment of the second point cloud and its corresponding line parameters (pj, vj), a directed bounding box is constructed using the direction vector vj as the principal axis direction of the OBB. An OBB is a cuboid bounding box that can tightly enclose a point cloud in any direction, typically constructed based on PCA (Principal Component Analysis). By performing PCA analysis on the point cloud data, the three principal directions of the point cloud distribution (i.e., the eigenvectors of the covariance matrix) are calculated, and these are used as the three axis directions of the bounding box. In this embodiment, one principal axis of the directed bounding box is forced to align with the fitted reinforcement direction vector vj, making the length direction of the bounding box consistent with the extension direction of the reinforcement. After determining the principal axis direction, the point cloud is projected onto two orthogonal directions perpendicular to this principal axis, and the projection range is calculated to determine the width and height of the bounding box, thereby constructing a complete directed bounding box.

[0115] S702. Determine the central axis of the oriented bounding box based on the position information in the straight line parameters corresponding to each steel bar segment and the bending point.

[0116] It should be noted that the obtained straight line parameters include the three-dimensional coordinate vectors of known points on the straight line equation of each rebar segment. For adjacent rebar segments with bends, the coordinates of the bends are determined by the intersection constraint mechanism. The parameterized form of this central axis in space is: l : p = p i +τ · v .

[0117] S703. Calculate the cross-sectional radius of the oriented bounding box in the direction perpendicular to the main axis.

[0118] It should be noted that the cross-section of OBB perpendicular to the principal axis is a rectangle, and the dimensions of this rectangle reflect the distribution range of the second point cloud perpendicular to the extension direction of the reinforcing bar. Since the cross-section of the reinforcing bar is usually circular, this rectangular cross-section is equivalent to a circle, and its radius is the cross-sectional radius of the reinforcing bar. The cross-sectional radius is calculated as follows: project all points in the second point cloud onto a plane perpendicular to the principal axis vj, calculate the distribution range of the projected points on the plane, and take the average of the semi-axial lengths in two orthogonal directions as the cross-sectional radius.

[0119] S704. Using the central axis as the center line of the reinforcing bar and the cross-sectional radius as the radius of the reinforcing bar, generate the solid model of the reinforcing bar corresponding to each reinforcing bar segment.

[0120] It should be noted that: using the central axis of the OBB as the centerline of the reinforcing bar and the cross-sectional radius as the reinforcing bar radius, a solid model of the cylindrical reinforcing bar corresponding to each segment is generated. This cylinder can be represented as the set of all points in space whose distance from the central axis is less than or equal to the cross-sectional radius. The length of this cylinder is the actual length of the reinforcing bar segment, determined by adjacent bends or the ends of the reinforcing bars. For continuous reinforcing bars with bends, adjacent cylinders are joined end-to-end at the bends to form a complete solid model of the reinforcing bar. The principle of the inscribed cylinder in OBB is as follows: Figure 13 As shown.

[0121] In this embodiment, a oriented bounding box is constructed with the fitted direction vector as the main axis, tightly enclosing the second point cloud. This accurately reflects the spatial orientation and cross-sectional dimensions of the reinforcing bars, effectively avoiding dimensional estimation errors caused by inconsistencies between the bounding box direction and the reinforcing bars. Furthermore, by directly using the fitted linear parameters as the central axis of the OBB, a seamless transition from linear fitting to solid modeling is achieved, ensuring the continuity of the modeling process and the accuracy of parameter transfer. Further, the diameter parameters of the reinforcing bars are automatically obtained by calculating the cross-sectional radius of the OBB perpendicular to the main axis, eliminating the need for manual measurement or prior input, thus achieving fully automated extraction of the reinforcing bar geometric parameters. Finally, a cylindrical solid model is generated using the central axis as the reinforcing bar centerline and the cross-sectional radius as the reinforcing bar radius. This transforms the abstract linear parameters and radius values ​​into a concrete three-dimensional geometric entity, providing an accurate geometric model for subsequent finite element mesh generation and structural analysis.

[0122] In some embodiments of the present invention, generating the finite element model of the prefabricated reinforced concrete component based on the steel reinforcement solid model and the concrete solid model includes: The steel reinforcement solid model is embedded into the concrete solid model to form the finite element model; the steel reinforcement solid model and the concrete solid model interact with each other through finite element contact constraints or embedded constraints; the volume mesh of the finite element model is a tetrahedral volume mesh.

[0123] It should be noted that: through the above embodiments, AABB bounding box intersection tests and surface-line-point topological reasoning were performed on each concrete plane to determine the closed boundary of the concrete (i.e., the concrete solid model). Then, through the above embodiments, an OBB inscribed cylinder was constructed for each segment of the second point cloud using the fitted direction vector as the main axis to obtain the geometric model of the reinforcing steel (i.e., the solid model of the reinforcing steel). Subsequently, the concrete solid model and the solid model of the reinforcing steel were combined into a closed three-dimensional entity. Specifically, the concrete solid model was retained as a complete entity, and the cylindrical solid models of each reinforcing steel segment were placed one by one in the corresponding spatial positions inside the concrete solid model according to the bending point position and the direction of the central axis. The two exist independently and do not subtract from each other. That is, the closed three-dimensional entity in this invention refers to the solid model of the reinforcing steel being embedded as an independent entity inside the concrete, rather than deducting the volume occupied by the reinforcing steel from the concrete to form a hollow entity. The two achieve mechanical interaction through finite element contact constraints or embedded constraints.

[0124] Finite element contact constraints define contact pairs at the interface between the steel reinforcement and concrete, transferring the normal pressure and tangential friction between them through a contact algorithm. Embedded constraints embed the steel reinforcement solid model into the concrete solid model, establishing a correlation between the degrees of freedom of the steel reinforcement nodes and the degrees of freedom of the concrete elements through constraint equations, thereby achieving deformation coordination and load transfer between the two. Both constraint methods can effectively simulate the bond-slip relationship and collaborative working mechanism between the steel reinforcement and concrete.

[0125] Furthermore, the closed 3D solid model formed by the above combination is tetrahedral meshed to generate a tetrahedral volume mesh for the finite element model. Specifically, tetrahedral meshing is performed using the pygmsh library, with a maximum mesh size set to hmax = 50 mm. During meshing, the concrete solid model and the reinforced concrete solid model are meshed separately. The mesh nodes at their contact interface can be inconsistent (non-matched meshes), and the mechanical connection between them is established through contact constraints or embedded constraints. The final output volume mesh is in STEP format, compatible with commercial finite element software such as Abaqus / ANSYS.

[0126] In this embodiment, by embedding the steel reinforcement solid model as an independent entity within the complete concrete solid model, the risk of failure in complex geometric Boolean operations is effectively avoided, significantly simplifying the difficulty and computational cost of entity combination and mesh generation. Furthermore, by implementing contact constraints or embedded constraints to realize the interaction between the steel reinforcement and concrete, the collaborative working state and interface mechanical behavior of the two during the stress process can be accurately simulated, resulting in higher analysis accuracy compared to modeling the steel reinforcement and concrete as a single unit. Furthermore, by directly generating a tetrahedral volume mesh and outputting it in a format compatible with commercial finite element software, the generated finite element model can be directly imported into software such as Abaqus and ANSYS for structural analysis without manual conversion, saving time and labor costs associated with model conversion. The entire finite element model generation process achieves end-to-end automation from 3D point cloud data to a computable finite element model, effectively improving the efficiency and accuracy of preprocessing for pre-analysis of prefabricated reinforced concrete structural components.

[0127] To better implement the prefabricated reinforced concrete component model generation method in this embodiment of the invention, based on the prefabricated reinforced concrete component model generation method, correspondingly, as follows: Figure 8 As shown, this embodiment of the invention also provides a prefabricated reinforced concrete component model generation system 800, which includes: Point cloud acquisition module 801 is used to acquire three-dimensional point cloud data of prefabricated reinforced concrete components; The semantic segmentation module 802 is used to perform clustering and segmentation on the three-dimensional point cloud data based on the geometric features extracted from the three-dimensional point cloud data to obtain a first point cloud corresponding to concrete and a second point cloud corresponding to steel bar segments. The fitting module 803 is used to perform multi-plane fitting on the first point cloud to obtain the plane equations corresponding to each concrete; to perform spatial polyline fitting on the second point cloud to obtain the linear parameters corresponding to each steel bar segment, and to make the fitting lines corresponding to adjacent steel bar segments intersect under preset constraints to obtain bending points. The model generation module 804 is used to construct a closed concrete solid model based on the plane equation, and to construct a steel reinforcement solid model based on the straight line parameters and the bending points; and to generate a finite element model of the prefabricated reinforced concrete component based on the steel reinforcement solid model and the concrete solid model.

[0128] It should be noted that a precast reinforced concrete composite beam was used as the test object. This component has complex geometric features such as dense reinforcement, small local surfaces, and concave surfaces. Gaussian noise with a standard deviation of 2 mm was added to simulate real engineering scanning conditions. The processing parameters were set as follows: neighborhood search radius r = 40 mm, preset threshold of 3 mm (considering noise increase), maximum number of fitted planes Nmax = 20, and the density clustering neighborhood radius was set according to the actual reinforcement diameter. In this embodiment, the number of clusters k=2 in K-Means (K-Means Clustering) was used to divide the point cloud into high flatness clusters and high linearity clusters.

[0129] The following combination Figures 10 to 14 The implementation process and intermediate results of the technical solution of the present invention will be described in detail. For example... Figure 10 As shown, this demonstrates the complete automated generation process from pre-processed point clouds of prefabricated reinforced concrete components to the final finite element mesh. Figure 10 The processing logic is divided into four steps from top to bottom, with each step separated by a dashed box and arrows indicating the direction of data flow.

[0130] Figure 10 The first step is the primary clustering of concrete and steel reinforcement point clouds (i.e., the first and second point clouds). Figure 10 The first step, inputting from the left, is the pre-processed point cloud of the precast components. It displays a complete precast composite beam in gray point cloud form, showing the point cloud distribution of multiple parallel reinforcing bars within it. After neighborhood feature extraction, the output is a visualization of eight neighborhood geometric features, arranged in two rows and four columns. The first row, from left to right, shows linearity, flatness, sphericity, and total variance; the second row, from left to right, shows anisotropy, information entropy, tortuosity, and curvature. The feature values ​​are indicated by varying shades of red, visually demonstrating the ability of different geometric features to distinguish between reinforcing bars and concrete. A dark horizontal bar at the bottom indicates the K-means algorithm, showing that based on the aforementioned neighborhood features, the K-means clustering algorithm initially divides the point cloud into two classes.

[0131] Figure 10 The second step is to fit the concrete surface. Figure 10 The second step involves the left-hand input being clustered and identified as the first point cloud (i.e., the point cloud corresponding to the concrete), which is displayed in light green. Figure 10The second step demonstrates the iterative extraction process of the RANSAC algorithm's multi-plane fitting, labeled as six planes, corresponding to the results of six iterations. The specific process of each iteration is as follows: three non-collinear points are randomly sampled from the remaining first point cloud. A candidate plane equation is calculated from these three points. Then, the perpendicular distance from all points in the point cloud to this candidate plane is calculated. The number of points whose distance is less than a set threshold is counted; these points are called interior points of the plane. This random sampling and plane calculation process is repeated multiple times, and finally, the plane with the most interior points is retained as the valid plane fitted in this iteration. Figure 10 The red portion in the second step represents the points within the concrete plane fitted in the current iteration, signifying a successfully extracted principal concrete plane. After each plane is extracted, its corresponding interior points are removed from the overall point cloud. The same RANSAC fitting process is then repeated on the remaining point cloud, sequentially extracting the second, third, and so on up to the sixth plane, thus gradually separating each principal surface from the first point cloud. The right side shows the first point cloud after fitting, using various colors such as purple, yellow, green, and cyan to distinguish different planes. Each color corresponds to a fitted concrete surface, indicating that all concrete planes of the precast component have been successfully identified, and each plane has obtained its corresponding plane equation parameters, laying the foundation for subsequent topology reconstruction and boundary reasoning.

[0132] Figure 10 The third step is the identification of three-dimensional line segments of steel bars, which belongs to the second stage of point cloud purification and skeleton fitting. Figure 10 The arrow pointing from the second step output to the third step indicates that non-reinforced points are removed based on the plane equation. This means that the candidate point cloud output from the first step clustering is purified by using the concrete plane equation obtained from the second step to remove incorrectly segmented concrete boundary points. Figure 10 The third step involves inputting candidate point clouds that have been identified as the second point cloud (i.e., the point cloud corresponding to the rebar segment) after clustering. These point clouds are then superimposed with the purification information from the second step. After purification, a pure second point cloud is obtained, which is then processed by the adjusted RANSAC algorithm. Figure 10 The third step outputs all the points on the left after segmentation, showing the complete point cloud of the composite beam. The concrete part is colored, and the steel reinforcement skeleton is embedded in it in the form of multi-colored straight line segments, which intuitively presents the spatial relative position of the steel reinforcement and concrete.

[0133] Figure 10 The fourth step is topology reconstruction and volume mesh generation, which belongs to the three-dimensional solid reconstruction and finite element mesh generation stage. Figure 10 The fourth step, marked on the left, shows the intersection of surfaces into lines and lines into points, demonstrating the closed concrete entity obtained by reasoning through the intersection of various concrete planes. The surface has been divided into a triangular mesh. Figure 10The fourth step, intermediate annotation, solves the inscribed cylinder of the OBB bounding box, showing the spatial wireframe model of multiple reinforcing bars. It indicates that a directed bounding box is constructed for each reinforcing bar with the fitted direction vector as the main axis, and the geometric model of the inscribed cylinder as the reinforcing bar is calculated. Figure 10 The fourth step, as shown on the right, displays the volume mesh in Vista, showcasing the final generated complete finite element model. The red, semi-transparent concrete solid contains multiple gray steel cylindrical reinforcements, and the overall mesh is a tetrahedral structure, ready for direct structural analysis. In summary... Figure 10 The three-dimensional laser scanning point cloud data of prefabricated reinforced concrete components is obtained as input. Then, through improved K-Means unsupervised semantic segmentation based on neighborhood feature engineering, the point cloud is clustered into high flatness clusters and high linearity clusters. The concrete plane equation is used for secondary purification to obtain the first point cloud (corresponding to concrete) and the second point cloud (corresponding to steel bar segments). Next, RANSAC multiplane fitting was performed on the first point cloud to extract the equations of each concrete plane. DBSCAN density clustering was performed on the second point cloud to separate the point cloud clusters of individual steel bars. Enhanced RANSAC spatial polyline fitting was then performed on each steel bar point cloud cluster to obtain the straight line parameters and bending point coordinates of each steel bar segment. Finally, AABB bounding box intersection test and surface-line-point topology inference were performed on each concrete plane to determine the closed concrete solid model. For each segment of the second point cloud, an OBB inscribed cylinder was constructed with the fitted direction vector as the main axis to obtain the steel bar solid model. The concrete solid and the steel bar solid were combined into a closed three-dimensional solid, and tetrahedral meshing was performed using the pygmsh library. The maximum mesh size hmax=50mm was set, and the output was in STEP format compatible with commercial finite element software such as Abaqus / ANSYS, thus generating the finite element model of the prefabricated reinforced concrete component.

[0134] Figure 11 Comparison of point cloud semantic segmentation results for prefabricated reinforced concrete components. Figure 11 (a) in the image is the original point cloud of the complete precast composite beam, which is gray in color and shows a complete beam member with end bars and top stirrups. Figure 11 The equal sign below (a) indicates that the complete point cloud can be decomposed into two parts: steel reinforcement and concrete. Figure 11 (b) in the image is the second point cloud (reinforcement point cloud) of the coarse classification, which is orange-red and shows the point cloud distribution of all the reinforcement bars in the component, including the top stirrups and the bottom main reinforcement extensions. Figure 11 (b) and Figure 11 Add a plus sign between (c) in the text. Figure 11 (c) in the diagram represents the first point cloud (concrete point cloud), which is light green and shows the point cloud portion of the concrete matrix within the component. Together, these three elements illustrate that the complete point cloud of the component is composed of the superposition of the second and first point clouds. Figure 11 (d) in Figure 11(e) in Figure 11 (f) in the figure shows the visualization effects of three different geometric features on the point cloud. Figure 11 (d) in Figure 11 (e) in Figure 11 In the (f) function, the eigenvalue is represented by a color map. The closer the color is to red, the higher the eigenvalue is, and the closer it is to blue, the lower the eigenvalue is. Figure 11 (d) in the figure represents the visualization results of the linearity feature. The second point cloud region is clearly red, while the first point cloud region is blue-green, which intuitively reflects the high response characteristics of the linearity feature to linear structures such as steel bars. The linearity value of the second point cloud region is significantly higher than that of the first point cloud region. Figure 11 (e) in the figure shows the visualization results of the flatness feature. The first point cloud surface area is presented in a warm reddish hue, while the second point cloud area is presented in blue, demonstrating the ability of the flatness feature to distinguish planar structures such as concrete. The flatness value of the first point cloud surface area is significantly higher than that of the second point cloud area. Figure 11 (f) in the diagram shows the visualization results of the sphericity feature. The first and second point cloud regions exhibit different color distributions, with the first region leaning towards green and the second towards blue, demonstrating the different responses of the sphericity feature to the two types of point clouds. This provides another geometric dimension to aid in distinguishing between steel reinforcement and concrete. Thus, in the semantic segmentation stage, by extracting the 7-dimensional geometric features of each point in the point cloud (including linearity, flatness, sphericity, total variance, anisotropy, feature entropy, and curvature variation), the normalized coordinates are concatenated with these features to form a 10-dimensional augmented feature vector. The K-Means algorithm (with a cluster size of k=2) using an improved distance metric is employed to cluster the point cloud into high flatness clusters and high linearity clusters. Then, the concrete plane equation obtained through subsequent RANSAC fitting is used for secondary purification to remove mis-segmented points, ultimately yielding a clean first and second point cloud. Figure 11 As shown, the light-colored dot cloud represents concrete surface points, while the dark-colored dot cloud represents steel reinforcement points. After the above processing, the two can be accurately distinguished.

[0135] Figure 12This image shows the point cloud segmentation and rebar identification results for prefabricated reinforced concrete components. In the geometric primitive fitting stage, RANSAC multiplane fitting is performed on the first point cloud, iteratively extracting up to Nmax=20 planes and obtaining the equations for each plane. For the second point cloud, DBSCAN density clustering is first performed, setting the neighborhood radius according to the requirement in the "Code for Design of Concrete Structures" that the rebar spacing should not be less than 1.5 times the diameter, separating individual rebar point cloud clusters. Then, enhanced RANSAC spatial polyline fitting is performed on each cluster. In this fitting process, the first round of RANSAC obtains the longest straight line segment. Subsequent iterations introduce an intersection constraint mechanism—by constructing a constraint plane and projecting random sampling points onto this plane, the corrected subsequent straight lines are forced to intersect precisely with the first round straight lines at the bend points, ensuring that adjacent straight line segments intersect precisely at spatial bend points, until the average linearity of the remaining point cloud is lower than a threshold. t L The process terminates when the value reaches 0.7. Figure 12 (a) in the image represents the first point cloud after segmentation (concrete point cloud), which is generally shaped like a long, narrow beam. Figure 12 In (a), the left end of the component has multiple areas of different colors, including green, purple, and cyan, which correspond to the various vertical and concave surfaces of the concrete end, respectively. This visually demonstrates that after semantic segmentation, the concrete surfaces have been accurately identified and distinguished by different colors. Figure 12 (b) shows the identification results of the voxel downsampling normalized point cloud and the second point cloud. The voxel downsampling normalized point cloud represents the homogenized point cloud data after voxel downsampling preprocessing. The identified second point cloud is displayed in multiple different colors, marked as the identified second point cloud, with different colors representing different reinforcing bars. Figure 12 In section (b), there is a row of U-shaped stirrups at the top, arranged sequentially, with each stirrup distinguished by a unique color. Figure 12 In section (b), there are two protruding main reinforcing bars at the bottom left end, also marked with different colors. For example... Figure 12 As shown, the spatial layout of the concrete plane boundary and the steel reinforcement skeleton is clearly visible, indicating that after DBSCAN density clustering, each individual steel bar has been successfully separated and identified, each steel bar has an independent color mark, and each plane and steel bar segment has been accurately fitted.

[0136] like Figure 13 As shown, a large number of black scattered points are distributed in the three-dimensional grid coordinate system, representing the point cloud data of a single steel bar. The overall distribution is in the shape of a vertical cylinder, which intuitively shows the spatial morphology of the second point cloud. Figure 13 A black vertical line running through the center of the point cloud (extending along the Z-axis) represents the central axis of the reinforcing bar obtained by enhanced RANSAC fitting. This line is the principal axis direction of the reinforcing bar. Figure 13The cuboid wireframe that fits the outer contour of the point cloud is an OBB (Directed Bounding Box). The direction of the OBB is along the straight line of the rebar axis (not the coordinate axis), and it is a directed bounding box constructed based on the actual spatial orientation of the rebar. In the rebar solid modeling stage, for each segment of the second point cloud, the direction vector v obtained by RANSAC fitting is... j Construct an OBB (directed bounding box) along the main axis direction, and calculate the OBB along the axis perpendicular to v. j The cross-sectional radius in the direction is used as the diameter of the reinforcing bar, and the OBB central axis is used as the center line of the reinforcing bar. The complete geometric parameters of the reinforcing bar are obtained by combining the coordinates of the bending point, thereby transforming the abstract point cloud data into a specific cylindrical geometric model, providing an accurate geometric representation of the reinforcing bar for subsequent finite element mesh generation.

[0137] Figure 14 The diagram shows the concrete boundary topology reconstruction process and the final volume mesh result. Figure 14 The diagram is divided into three sub-graphs arranged from top to bottom, showing the boundary defects, the corrected boundary, and the complete mesh model in sequence. Figure 14 In the image (a), the concrete boundary frame obtained from the initial topological inference is shown, with the overall outline of the composite beam delineated in red. Figure 14 The shape of two concave holes at the left end of (a) indicates that some edges are not closed due to defects in the intersection of straight lines. This visually demonstrates the boundary non-closure problem in the initial surface-line-point topological reasoning, that is, the endpoints of some intersecting lines fail to intersect accurately, resulting in gaps in the boundary of the concrete surface. Figure 14 (b) in the image represents the concrete boundary frame after progressively expanding the threshold correction. (and) Figure 14 (a) in the comparison, Figure 14 In (b), the boundaries of the two concave holes on the left end are completely closed, and all the endpoints of the intersection lines intersect accurately, forming a complete and closed concrete surface profile. Figure 14 (b) in the paper demonstrates how a strategy of gradually relaxing the intersection determination threshold can be used to successfully repair boundary defects and obtain a closed concrete geometric boundary. Figure 14 (c) in the figure represents the final generated complete finite element mesh model. Figure 14 In (c), the concrete part is shown in red, and the surface of the red area is covered with triangular grid lines, representing the surface of the tetrahedral grid. Figure 14 In (c), there is a row of U-shaped stirrups on the top of the concrete (embedded above the concrete in the form of a gray line frame). Figure 14 In (c), there are two protruding main ribs at the bottom left end (marked in gray). Figure 14 (c) shows the complete finite element model obtained by tetrahedral meshing after combining the concrete boundary and the reinforcing steel skeleton, which can be directly used for structural analysis calculations. In the mesh generation stage, such as... Figure 14As shown, the concrete solid model and the steel reinforcement solid model are combined into a closed three-dimensional solid (the steel reinforcement is embedded inside the concrete as an independent entity, and the two interact through finite element contact or embedded constraints). Tetrahedral meshing is performed using the pygmsh library, with a maximum mesh size hmax=50mm. The output is in STEP format, compatible with commercial finite element software such as Abaqus / ANSYS, thus generating a finite element model of the prefabricated reinforced concrete component. This volumetric mesh model can be directly imported into commercial finite element software for structural analysis, achieving fully automated generation from point cloud to finite element model.

[0138] The volumetric mesh model was imported into ABAQUS using the STEP format for validation. Material parameters used were C30 concrete (JGJ55-2011) and HRB400 steel reinforcement (GB / T 1499.2-2018). Boundary conditions were fixed-end constraints, with uniformly distributed pressure applied to the top surface. Cantilever beam compression validation results: simulated displacement 3.74 mm, theoretical value 3.68 mm, error approximately 1.6%; simulated stress 10.54 MPa, theoretical value 10.50 MPa, error approximately 0.4%. Furthermore, Gaussian noise with a standard deviation of 0 mm to 12 mm was added to the original point cloud. The complete workflow was executed at each noise level, and evaluation metrics included the integrity of the concrete surface reconstruction, the accuracy of steel reinforcement positioning, and the error in steel reinforcement diameter reconstruction. Results show that when the standard deviation does not exceed 1 cm, the method can still complete the full 3D reconstruction—all principal planes of the concrete surface can be detected and fitted, the steel reinforcement positioning accuracy remains stable, and the reconstructed steel reinforcement diameter increases slightly with noise but remains within acceptable limits. This noise resistance (1 cm standard deviation) exceeds the accuracy requirements of typical 3D laser scanners (±1.5 mm to ±2 mm). When the point cloud quality is poor (e.g., excessive noise or severe missing data), reducing the Nmax value (e.g., setting it to 6) will cause the algorithm to abandon fitting small concave surfaces and local details, prioritizing the reconstruction of the main structure (main concrete planes and main reinforcement).

[0139] The prefabricated reinforced concrete component model generation system 800 provided in the above embodiments can realize the technical solutions described in the above embodiments of the prefabricated reinforced concrete component model generation method. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the prefabricated reinforced concrete component model generation method, which will not be repeated here.

[0140] like Figure 9 As shown, the present invention also provides an electronic device 900. The electronic device 900 includes a processor 901, a memory 902, and a display 903. Figure 9 Only some components of the electronic device 900 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0141] In some embodiments, processor 901 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 902 or process data, such as the prefabricated reinforced concrete component model generation method of the present invention.

[0142] In some embodiments, processor 901 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 901 may be local or remote. In some embodiments, processor 901 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, etc., or any combination thereof.

[0143] In some embodiments, memory 902 may be an internal storage unit of electronic device 900, such as a hard disk or memory of electronic device 900. In other embodiments, memory 902 may also be an external storage device of electronic device 900, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 900.

[0144] Furthermore, the memory 902 may include both internal storage units of the electronic device 900 and external storage devices. The memory 902 is used to store application software and various types of data installed on the electronic device 900.

[0145] In some embodiments, display 903 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light Emitting Diode) touchscreen. Display 903 is used to display information from electronic device 900 and to display a visual user interface. Components 901-903 of electronic device 900 communicate with each other via a system bus.

[0146] In one embodiment, when processor 901 executes the prefabricated reinforced concrete component model generation program in memory 902, the following steps can be implemented: Obtain 3D point cloud data of prefabricated reinforced concrete components; Based on the geometric features extracted from the three-dimensional point cloud data, the three-dimensional point cloud data is clustered and segmented to obtain the first point cloud corresponding to the concrete and the second point cloud corresponding to the steel bar segment. Multiplane fitting is performed on the first point cloud to obtain the plane equations corresponding to each concrete. Spatial polyline fitting is performed on the second point cloud to obtain the linear parameters corresponding to each steel bar segment, and the fitted lines corresponding to adjacent steel bar segments intersect under preset constraints to obtain bending points. A closed concrete solid model is constructed based on the plane equation, and a steel reinforcement solid model is constructed based on the straight line parameters and the bending points. Based on the steel reinforcement solid model and the concrete solid model, a finite element model of the prefabricated reinforced concrete component is generated.

[0147] It should be understood that when the processor 901 executes the prefabricated reinforced concrete component model generation program in the memory 902, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0148] Furthermore, this embodiment of the invention does not specifically limit the type of electronic device 900 mentioned. Electronic device 900 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the invention, electronic device 900 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0149] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0150] The above provides a detailed description of the method, system, and equipment for generating prefabricated reinforced concrete component models provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for generating a model of a prefabricated reinforced concrete component, characterized in that, include: Acquire 3D point cloud data of prefabricated reinforced concrete components; Based on the geometric features extracted from the three-dimensional point cloud data, the three-dimensional point cloud data is clustered and segmented to obtain the first point cloud corresponding to the concrete and the second point cloud corresponding to the steel bar segment. Multiplane fitting is performed on the first point cloud to obtain the plane equations corresponding to each concrete. Spatial polyline fitting is performed on the second point cloud to obtain the linear parameters corresponding to each steel bar segment, and the fitted lines corresponding to adjacent steel bar segments intersect under preset constraints to obtain bending points. A closed concrete solid model is constructed based on the plane equation, and a steel reinforcement solid model is constructed based on the straight line parameters and the bending points. Based on the steel reinforcement solid model and the concrete solid model, a finite element model of the prefabricated reinforced concrete component is generated.

2. The method according to claim 1, characterized in that, The step of clustering and segmenting the three-dimensional point cloud data based on the geometric features extracted from the data to obtain a first point cloud corresponding to the concrete and a second point cloud corresponding to the steel reinforcement segment includes: Obtain the set of neighborhood points within the neighborhood range of each point in the three-dimensional point cloud data; Calculate the neighborhood centroid of each point based on the neighborhood point set, and construct the covariance matrix of each point based on the neighborhood centroid and the neighborhood point set; The covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvalues; Calculate the linearity, flatness, sphericity, total variance, anisotropy, feature entropy, and curvature change of each point in the three-dimensional point cloud data based on the feature values. After normalizing the spatial coordinates of each point, they are concatenated with the linearity, flatness, sphericity, total variance, anisotropy, feature entropy, and curvature change to form a combined feature vector; Based on the combined feature vector, the points in the three-dimensional point cloud data are clustered to divide the point cloud into high flatness clusters and high linearity clusters, thereby obtaining the first point cloud and the candidate point cloud corresponding to the steel bar segment. The candidate point cloud is filtered using the plane equation obtained by multi-plane fitting. Points in the candidate point cloud whose distance to the nearest concrete plane is less than a preset threshold are removed from the candidate point cloud to obtain the second point cloud.

3. The method according to claim 1, characterized in that, The step of performing multi-plane fitting on the first point cloud to obtain the plane equations corresponding to each concrete element includes: Three non-collinear points are randomly sampled from the first point cloud, and the candidate plane equation is calculated based on the sampled three non-collinear points; Count the number of target points in the first point cloud whose distance to the candidate plane equation is less than a preset threshold; The process of iteratively performing the random sampling, calculating the candidate plane equation, and counting the number of target points is used, and the candidate plane equation with the largest number of points is retained as the final plane equation. Remove points whose distance to the final plane equation is less than a preset threshold from the first point cloud. Repeat the above process for the remaining point cloud until the preset maximum number of fitted planes is reached or the number of remaining points in the first point cloud is less than a preset threshold, and obtain the plane equation corresponding to each concrete.

4. The method according to claim 1, characterized in that, The step of performing spatial polyline fitting on the second point cloud to obtain the linear parameters corresponding to each steel bar segment includes: A density clustering algorithm is used to cluster the second point cloud with a pre-set neighborhood radius to obtain multiple single second point cloud clusters; the neighborhood radius is set according to the diameter of the reinforcing bar. Spatial polyline fitting is performed on each of the single second point cloud clusters to obtain the linear parameters corresponding to each steel bar segment.

5. The method according to claim 1, characterized in that, The step of making the fitted straight lines corresponding to adjacent steel bar segments intersect under preset constraints to obtain the bending point includes: Obtain the direction vector of the first fitted line corresponding to the steel bar segment in the first round of fitting, as well as the known points of the first fitted line; Obtain the direction vector of the straight line to be fitted corresponding to the steel bar segment adjacent to the steel bar segment to be fitted in the first round of fitting; The normal vector is determined based on the direction vector of the first fitted line and the direction vector of the line to be fitted; The plane with the normal vector as its normal direction and passing through the known point is used as the constraint plane; The random sampling points of the line to be fitted are projected onto the constraint plane to obtain the projection points; Using the projection point as the starting point and the direction vector of the line to be fitted as the direction, a corrected line to be fitted is generated. The bending point is obtained from the intersection of the corrected line to be fitted and the first fitted line.

6. The method according to claim 1, characterized in that, The construction of a closed concrete solid model based on the plane equation includes: Determine the plane corresponding to each concrete according to the plane equation corresponding to each concrete, and construct an axisymmetric bounding box for each plane respectively. An intersection test is performed on each of the axisymmetric bounding boxes. Based on the intersection test results, the intersecting plane pairs are determined, and the intersection line between the intersecting plane pairs is calculated. The endpoints of each intersection line are determined according to a preset threshold sequence, and the boundaries of each plane are formed by the intersection lines and their endpoints. The concrete solid model is formed based on the boundaries of each plane.

7. The method according to claim 1, characterized in that, The construction of the steel reinforcement solid model based on the straight line parameters and the bending points includes: A oriented bounding box is constructed using the direction vector in the linear parameters corresponding to each steel bar segment as the main axis direction. The central axis of the oriented bounding box is determined based on the position information in the straight line parameters corresponding to each steel bar segment and the bending point. Calculate the cross-sectional radius of the directed bounding box in the direction perpendicular to the principal axis; Using the central axis as the center line of the reinforcing bar and the cross-sectional radius as the radius of the reinforcing bar, a solid model of the reinforcing bar corresponding to each reinforcing bar segment is generated.

8. The method according to claim 1, characterized in that, The process of generating the finite element model of the prefabricated reinforced concrete component based on the steel reinforcement solid model and the concrete solid model includes: The steel reinforcement solid model is embedded into the concrete solid model to form the finite element model; the steel reinforcement solid model and the concrete solid model interact with each other through finite element contact constraints or embedded constraints; the volume mesh of the finite element model is a tetrahedral volume mesh.

9. A prefabricated reinforced concrete component model generation system, characterized in that, include: The point cloud acquisition module is used to acquire three-dimensional point cloud data of prefabricated reinforced concrete components. The semantic segmentation module is used to perform clustering and segmentation on the three-dimensional point cloud data based on the geometric features extracted from the three-dimensional point cloud data to obtain the first point cloud corresponding to the concrete and the second point cloud corresponding to the steel bar segment. The fitting module is used to perform multi-plane fitting on the first point cloud to obtain the plane equations corresponding to each concrete; to perform spatial polyline fitting on the second point cloud to obtain the linear parameters corresponding to each steel bar segment, and to make the fitting lines corresponding to adjacent steel bar segments intersect under preset constraints to obtain bending points. The model generation module is used to construct a closed concrete solid model based on the plane equation, and to construct a steel reinforcement solid model based on the straight line parameters and the bending points; and to generate a finite element model of the prefabricated reinforced concrete component based on the steel reinforcement solid model and the concrete solid model.

10. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the prefabricated reinforced concrete component model generation method according to any one of claims 1 to 8.

Images