Reverse bim modeling method, system and device for existing building space truss combined structure

Abstract

The application discloses a reverse BIM modeling method, system and equipment for a spatial rod system combined structure of an existing building, and belongs to the technical field of BIM modeling. In view of the fact that point cloud processing software in the prior art lacks special identification and reconstruction tools for a spatial rod system combined structure of a 'ball node-rod' type, and it is difficult to realize efficient and accurate reverse generation of a rod axis, the application accurately identifies the positions of ball nodes from measured point cloud data through curvature analysis, DBSCAN clustering, RANSAC spherical surface fitting and other technologies; and finally, a BIM model of the spatial rod system combined structure with high precision in line with engineering practice is generated according to the spatial constraint of the connecting line of the ball center of the ball node and the rod point cloud data and in combination with a two-stage geometric filtering strategy. The method effectively solves the problem that the measured point cloud data is discrete and difficult to process, and provides reliable technical support for health monitoring, digital filing and accurate modeling of the spatial rod system combined structure of the existing building.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of BIM modeling, and particularly relates to a method, system and device for reverse BIM modeling of a spatial rod system combined structure of an existing building. BACKGROUND

[0002] At present, existing buildings can only be managed by relying on two-dimensional design engineering drawings in operation and management, and for spatial rod system combined structures, it is difficult to manage the maintenance and health monitoring of existing buildings due to the complexity of three-dimensional space. The BIM model as a three-dimensional building information model can conveniently maintain and manage the spatial rod system combined structure and positioning, and it is difficult to quickly establish a three-dimensional BIM model of the existing building using existing engineering drawings.

[0003] With the development of technologies such as laser radar, point cloud has become the core carrier for obtaining three-dimensional information of objects, but the original point cloud has problems such as large data volume, much noise and information redundancy, and cannot directly extract the key rod (line segment) structure. However, the fields of building and civil engineering (steel structure health monitoring) urgently need to accurately identify the rod axis in the point cloud, and verify the results through visualization, quantify information (such as the number of rods, size deviation), save data to connect subsequent software (such as CAD, finite element analysis tools), etc. Especially for the spatial rod system combined structure of the existing building, due to the long-term service structure deformation and other monitoring needs, the generation of the rod axis of the spatial rod system combined structure becomes an important requirement for structure health monitoring. These rod axes of the spatial rod system combined structure can be used for structural engineering archiving (such as CAD drawing), structure BIM model generation, structure finite element mechanics analysis model establishment, etc., so the demand for the generation of the rod axis of the spatial rod system combined structure of the existing building is increasing.

[0004] However, the existing general point cloud processing software (such as CloudCompare, PCL, etc.) lacks special identification and reconstruction tools for "ball node-rod" type spatial structures, and users often need to perform a large amount of manual processing and secondary development of algorithms, which makes it difficult to achieve efficient and accurate reverse generation of rod axes. SUMMARY

[0005] In view of the above problems, the present application aims to provide a kind of existing building space pole system combined structure reverse BIM modeling method, system and equipment, by curvature analysis, DBSCA clustering, RANSAC spherical surface fitting etc.techniques, accurately identify the spherical node position from measured point cloud data;Again according to the space constraint of the spherical node's spherical center connecting line and pole piece point cloud data, combined with two-stage geometric filtering strategy, finally generate the BIM model of the space pole system combined structure in accordance with engineering practice high precision.This method effectively solves the problem that measured point cloud data is discrete and difficult to handle, and provides reliable technical support for existing building space pole system combined structure health monitoring, digitization filing and accurate modeling.

[0006] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0007] The present application provides an existing building space pole system combined structure reverse BIM modeling method, which comprises the following steps,

[0008] Obtain the point cloud data of the space pole system combined structure of the existing building, and determine the geometric center coordinates of the spherical node;

[0009] Based on point cloud matching degree filtering and spherical intersection filtering, determine the spatial axis of the pole piece between the spherical nodes, determine the spatial position and direction of the pole piece, and construct the two-dimensional model of the pole piece;

[0010] Extract the engineering drawing information of the space pole system combined structure of the existing building, transform the coordinate system of the two-dimensional model of the pole piece, and determine the spherical node information and pole piece information after coordinate system transformation;

[0011] Based on the spherical node information and pole piece information after coordinate system transformation, generate three-dimensional spherical node model and three-dimensional pole piece model, and perform BIM attribute mapping on the generated three-dimensional pole piece model to complete BIM modeling.

[0012] Further, the point cloud data of the space pole system combined structure of the existing building is obtained, and the geometric center coordinates of the spherical node include the following steps:

[0013] The space pole system combined structure of the existing building is scanned by a laser scanner, and the point cloud data of the space pole system combined structure is obtained;

[0014] The point cloud data of the space pole system combined structure is preprocessed by denoising;

[0015] From the preprocessed point cloud data, select the points with curvature greater than the threshold value as the spherical node candidate points;

[0016] Based on DBSCAN clustering method and RANSAC spherical surface fitting algorithm, fit the spherical surface according to the spherical node candidate points, and determine the geometric center coordinates of the spherical node.

[0017] Further, the DBSCAN clustering method and the RANSAC spherical surface fitting algorithm are used to fit a spherical surface according to the spherical node candidate points, and the geometric center coordinates of the spherical node are determined by the following steps:

[0018] DBSCAN clustering is performed on all the spherical node candidate points, and the spherical node candidate clusters are discretized, and the center point of each spherical node candidate cluster is determined;

[0019] For each spherical node candidate cluster, a spherical surface is fitted using the RANSAC spherical surface fitting algorithm, and the coordinates of the spherical center and the radius of the sphere are calculated;

[0020] According to the radius effective range and the distance deviation of all points in the spherical node candidate cluster to the spherical surface, the spherical surface fitting result is filtered, and the geometric center coordinates of the spherical node are determined.

[0021] Further, the point cloud matching degree filtering and the sphere intersection filtering are used to determine the spatial axis of the rod between the spherical nodes, determine the spatial position and direction of the rod, and construct a two-dimensional model of the rod, including the following steps:

[0022] Based on the geometric center coordinates of the spherical nodes, the distance between any two spherical nodes is calculated, and the effective node pairs are selected according to the distance between the spherical nodes to generate an initial rod axis candidate set;

[0023] Based on the point cloud matching degree filtering, the first candidate rod is selected from the initial rod axis candidate set;

[0024] Based on the sphere intersection filtering, the rod intersecting with three or more spherical nodes is removed from the first candidate rod, and the effective rod connecting only two spherical nodes is retained;

[0025] According to the spatial axis of the effective rod, the spatial position and direction of the rod are determined, and a two-dimensional model of the rod is constructed.

[0026] Further, the point cloud matching degree filtering is used to select the first candidate rod from the initial rod axis candidate set, which specifically includes the following steps:

[0027] For any candidate initial rod axis, the midpoint coordinates of the candidate initial rod axis are calculated;

[0028] Based on the KDTree index of the rod point cloud, the rod points within a 2-unit radius range around the midpoint of the candidate initial rod axis are extracted to form a rod point set;

[0029] For each rod point in the rod point set, the distance between the rod point and the candidate initial rod axis is calculated S n to the candidate initial rod axis ;

[0030] Set a distance threshold and count the number of pole points in the set. The number of pole points less than or equal to the distance threshold;

[0031] Set the dynamic filtering threshold to If the set of member points satisfies If the number of pole points less than or equal to the distance threshold is greater than the dynamic screening threshold, they are retained as first-level candidate poles; otherwise, they are discarded. is the line segment length of the candidate initial rod axis.

[0032] Furthermore, the extraction of engineering drawing information of the spatial frame structure of existing buildings, the transformation of the coordinate system of the two-dimensional model of the frame members, and the determination of the spherical node information and frame member information after the coordinate system transformation include the following steps:

[0033] Extract the engineering drawing information of the spatial pole system combination structure of the existing building, and transform the local coordinate system of the two-dimensional model of the pole to be consistent with the global coordinate system of the engineering drawing based on the ball node information, so as to obtain the ball node information after coordinate system transformation.

[0034] Based on the transformed local coordinate system, the members in the two-dimensional model are initially located. Then, the deviation of the members in the two-dimensional model is calculated and corrected to determine the member information after the coordinate system transformation.

[0035] Furthermore, the process of generating a 3D sphere node model and a 3D rod model based on the sphere node information and rod information after coordinate system transformation, and performing BIM attribute mapping on the generated 3D rod model to complete BIM modeling includes the following steps:

[0036] Establish a global BIM coordinate system and generate a 3D spherical node model based on the spherical node information after coordinate system transformation.

[0037] Based on the ball node position and the rod information after coordinate system transformation, determine the spatial orientation of the rod axis and the rod cross section, and generate a three-dimensional rod model;

[0038] Perform BIM attribute mapping on the generated 3D rod model to complete BIM modeling, and verify the accuracy.

[0039] On the other hand, the present invention also provides a reverse BIM modeling system for existing building spatial frame structures, used to implement the reverse BIM modeling method for existing building spatial frame structures as described above, the system comprising:

[0040] The sphere node positioning module is used to acquire point cloud data of the spatial pole system combination structure of existing buildings and determine the geometric center coordinates of the sphere nodes.

[0041] The rod positioning module determines the spatial axis of the rod between the spherical nodes based on point cloud matching degree filtering and sphere intersection filtering, determines the spatial position and direction of the rod, and constructs a two-dimensional model of the rod;

[0042] The coordinate transformation module is used for extracting engineering drawing information of the spatial rod system combined structure of the existing building, transforming the coordinate system of the two-dimensional model of the rod, and determining spherical node information and rod information after coordinate system transformation;

[0043] The BIM modeling module generates a three-dimensional spherical node model and a three-dimensional rod model based on the spherical node information and the rod information after the coordinate system transformation, and performs BIM attribute mapping on the generated three-dimensional rod model to complete BIM modeling.

[0044] In another aspect, the present application also provides an electronic device, which comprises at least one processor; and a memory connected in communication with the processor; wherein the memory stores instructions executed by the processor, and the instructions are executed by the processor to enable the processor to perform the reverse BIM modeling method of the spatial rod system combined structure of the existing building as described above.

[0045] The present application has the following advantages:

[0046] 1. The BIM modeling method in the present application accurately identifies the positions of the spherical nodes from the measured point cloud data through curvature analysis, DBSCAN clustering, RANSAC sphere fitting and other technologies; and finally generates a BIM model of the spatial rod system combined structure with high precision in accordance with the engineering practice, according to the spatial constraint of the connecting line of the spherical center of the spherical node and the rod point cloud data, and in combination with the two-stage geometric filtering strategy. This method effectively solves the problem of discrete and difficult processing of measured point cloud data, and reversely constructs a BIM model through measured point cloud data, thereby providing reliable technical support for health monitoring, digital filing and precise modeling of the spatial rod system combined structure of the existing building.

[0047] 2. The point cloud data acquisition method in the present application can efficiently and accurately collect the geometric feature information of the spatial rod system structure of the existing building, and realize rapid digital reconstruction of the structure form of the old building. Through automatic identification and association of the measured point cloud data and the spatial rod system combined structure components, the two-dimensional as-built drawing can be efficiently converted into an editable three-dimensional information model (BIM) in the absence of original design data, thereby effectively making up for the technical shortcoming of lack of spatial information in the maintenance work of the existing building, and improving the digital management and traceability of the building facilities.

[0048] 3、The method for obtaining the geometric center coordinates of the ball node in the application adopts the strategy of combining iterative optimization with multiple random consistency fitting (RANSAC), which can realize high-precision extraction of the center of the ball node under the condition of sparse point cloud, significantly reduce the dependence on high-density point cloud, and improve the calculation efficiency. In addition, through screening of multiple fitting results, the method effectively suppresses the interference of noise points and local defects on the results, ensuring the stability and robustness of the calculation of the geometric center coordinates of the ball node, and providing key technical support for the subsequent accurate identification of the rod and three-dimensional modeling.

[0049] 4、The method for generating the two-dimensional model of the rod in the application is based on the spatial matching relationship between the ball node geometric information and the pipeline point cloud, and realizes high-precision identification and geometric modeling of the spatial axis of the rod through the double constraint strategy of point cloud matching degree filtering and sphere intersection filtering. The method first calculates the distance between the ball nodes according to the geometric center coordinates of the ball nodes, generates an initial rod axis candidate set in combination with engineering constraints, and then uses KDTree search and point cloud projection algorithm to evaluate the spatial matching degree of the rod axis and the rod points, dynamically adjusts the threshold to adapt to the point cloud density of rod of different lengths, thereby effectively avoiding misidentification or missed identification. Subsequently, the invalid line segment intersecting with three or more ball nodes is removed through sphere intersection filtering, ensuring that the rod is connected to only two real ball nodes, and finally accurately reconstructing the spatial position and trend of the rod. This method not only significantly improves the geometric consistency and spatial accuracy of rod identification, but also realizes automatic topology generation from point cloud data to rod, greatly improving the BIM modeling efficiency and reliability of complex spatial rod combination structure. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The flowchart of the BIM modeling method based on the existing large-span spatial grid structure in the application.

[0051] Figure 2 The flowchart of the two-dimensional model generation of the existing large-span spatial grid structure in the application.

[0052] Figure 3 The flowchart of BIM modeling based on the two-dimensional model and engineering drawings in the application.

[0053] Figure 4 The point cloud data acquisition result in the simulation experiment of the application.

[0054] Figure 5 The BIM model finally constructed in the simulation experiment of the application. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the technical solutions of the application, the technical solutions of the application will be further described below in combination with the drawings and examples.

[0056] Embodiment One:

[0057] Referring to the accompanying drawings Figure 1 Embodiment One provides a reverse BIM modeling method for a spatial truss combined structure of an existing building. In this embodiment, the spatial truss combined structure is a spatial truss combined structure based on round steel pipe members, which includes spherical nodes and members. The method comprises the following steps,

[0058] Step 1: Obtain the point cloud data of the spatial truss combined structure of the existing building, and determine the geometric center coordinates of the spherical nodes;

[0059] Specifically, Step 1 comprises the following sub-steps,

[0060] Sub-step 101: Scan the spatial truss combined structure of the existing building by a laser scanner to obtain the point cloud data of the spatial truss combined structure;

[0061] More specifically, the spatial truss combined structure is divided into several layers as independent scanning units along the vertical direction (Z-axis direction). The number of layers and the height of each layer can be determined in combination with the height and complexity of the specific scanned spatial truss combined structure to ensure the relative independence of the round steel pipe members in each layer in space, facilitating subsequent scanning planning. Typically, it is divided into 3-5 layers. Taking three layers as an example: the bottom layer is 0-5m from the ground, focusing on scanning the spherical nodes and supports below the members; the middle layer is 5-15m, which is the main body area of the spatial truss combined structure, and the full length of the members should be covered; the top layer is above 15m, due to the high height, it should be scanned with climbing equipment (such as a lifting vehicle) to avoid missing the top spherical nodes.

[0062] Then, according to the distribution of the round steel pipe members in each layer, each layer is divided into several blocks along the X / Y axis (horizontal direction) to ensure that the scanning range of each laser scanner can cover 1-2 blocks, and the overlap rate of adjacent blocks is ≥20% to ensure accuracy.

[0063] Next, set up the laser scanner at a suitable position. The setting point of the laser scanner should meet the requirements that the laser beam emitted by the laser scanner can penetrate the gap of the round steel pipe members to cover all the round steel pipe members in the target block, and the horizontal rotation angle and vertical pitch angle can be adjusted to perform multi-angle scanning on the same target block to enhance the completeness of the point cloud data.

[0064] After each laser scanner is erected, appropriate scanning parameters are set according to the distance, material, etc. of the ball nodes and the bars, 3-5 blocks around are scanned first, and then the scanning units are scanned in turn. At the same time, the scanning time of a single laser scanner is controlled within 5-10 minutes (adjusted according to the size of the block), to ensure that the point cloud data has no fault. For each ball node, it needs to be scanned from at least two perpendicular directions (such as horizontal direction and 45° inclined direction), to ensure that the surface point cloud coverage rate of the ball node is ≥95%, and the scanning distance is controlled within 3-5m, the point cloud density is improved to 0.03m (point spacing), to ensure more accurate identification of high-curvature points. For bar scanning, at least one scanning section should be set every 2m along the length direction of the bar, to ensure that the bar has no "point cloud void"; for bars with different diameters, the scanning area is marked, and the curvature threshold can be adjusted in the program subsequently. Finally, the point cloud quality is checked in real time after scanning, including point cloud integrity, point cloud density, and control point accuracy. If data is missing or incorrect, it should be scanned in time, and finally the point cloud data of the spatial grid structure is formed.

[0065] Sub-step 102: denoising preprocessing of the point cloud data of the spatial truss combination structure;

[0066] More specifically, the denoising preprocessing of the point cloud data of the spatial truss combination structure includes voxel filtering and outlier filtering of the point cloud data of the spatial truss combination structure, to reduce the amount of point cloud data and remove noise points, and to prepare for the identification of ball nodes.

[0067] Optionally, based on the point cloud data of the spatial truss combination structure, a three-dimensional space is divided into a plurality of cubic voxels with an edge length of v For any cubic voxel, the n points P i ( x i , y i , z i ) ( i = 1, 2,..., n ), the center of the voxel is taken as the representative point of the cubic voxel. In each cubic voxel, the coordinates of the representative point P are represented as:

[0068] ;

[0069] In the formula, P i represents the i point in the cubic voxel, and x i , y i , z ​i ) is the coordinate of the i-th nearest neighbor of P in the cube voxel. i

[0070] For each P, the KDTree search is used to find the k nearest neighbors, and the Euclidean distance between P and each nearest neighbor is calculated. k d j j = 1, 2,..., k , and the average distance between P and the k nearest neighbors is calculated, denoted as the average neighborhood distance of P, denoted as: d j k

[0071] ;

[0072] In the formula, denotes the average neighborhood distance of P, j denotes the count of the nearest neighbors of P, k denotes the total number of nearest neighbors, is the coordinate of P; is the coordinate of the i-th nearest neighbor of P. j

[0073] The average value of the average neighborhood distance of P in all cube voxels is calculated, denoted as: μ

[0074] ;

[0075] In the formula, M denotes the total number of P, denotes the count of P.

[0076] If the average neighborhood distance of the representative point P in a certain cube voxel satisfies , then the point is determined as an outlier and needs to be removed, otherwise, the representative point P in the cube voxel is the valid point cloud. By indexing and filtering the valid point cloud, the outliers and the remaining points in the cube voxel corresponding to the outliers are removed, and the final output is the denoised point cloud data, which lays the foundation for curvature calculation. The valid P point set obtained after filtering is denoted as:

[0077] ;

[0078] Sub-step 103: From the pre-processed point cloud data, filter out points with curvature greater than a threshold as candidate points for spherical nodes.

[0079] ​​​​​​​For the denoised point cloud data, the neighborhood points of each point are searched using KDTree, and a neighborhood set is formed. The normal vector is estimated according to the neighborhood set. A neighborhood covariance matrix is constructed for the neighborhood set of any point, denoted as:

[0080] ;

[0081] wherein, B denotes the neighborhood covariance matrix, N denotes the number of neighborhood points in the neighborhood set, p denotes the count of neighborhood points, denotes the coordinate vector of the p th neighborhood point, denotes the centroid coordinates of all neighborhood points in the neighborhood set. is a column vector, is a row vector.

[0082] The neighborhood covariance matrix B is subjected to eigenvalue decomposition, and three non-negative eigenvalues are obtained, denoted as , and , and , the size of the eigenvalue reflects the dispersion degree of the neighborhood points in the corresponding eigenvector direction: the smallest, corresponding to the surface normal dispersion; and larger, corresponding to the surface tangent dispersion; define the curvature as the proportion of the smallest eigenvalue in the sum of the total eigenvalues (the higher the bending degree, the larger the relative proportion), then:

[0083] ;

[0084] From the calculated curvature results of any point, the points with a curvature greater than a threshold value (usually set to 0.015) are selected as the candidate points of the spherical node.

[0085] Sub-step 104: Based on the DBSCAN clustering method and the RANSAC spherical surface fitting algorithm, a spherical surface is fitted according to the candidate points of the spherical node, and the geometric center coordinates of the spherical node are determined.

[0086] More specifically, sub-step 104 can further include the following sub-steps:

[0087] Sub-step 1041: DBSCAN clustering is performed on all candidate points of the spherical node, and the candidate points are discretized into multiple spherical node candidate clusters. The center point of each spherical node candidate cluster is determined.

[0088] DBSCAN density clustering is performed on the ball node candidate points, all density-reachable points (i.e. non-core points with a neighborhood set of neighborhood points > 5) are classified into a cluster, and isolated points other than the above are removed, to finally obtain multiple discrete ball node candidate clusters. The center point of each ball node candidate cluster is calculated, i.e. the coordinate mean value of all points in each ball node candidate cluster, denoted as:

[0089] ;

[0090] In the formula, represents the center point coordinate of the ball node candidate cluster, K represents the number of points in the ball node candidate cluster, A is an arbitrary point in the ball node candidate cluster, is the coordinate of the point. The center point of the ball node candidate cluster serves as the initial reference position for subsequent spherical fitting. A

[0091] Sub-step 1042: for each ball node candidate cluster, a RANSAC spherical fitting algorithm is used to fit a sphere, to calculate the sphere center coordinates and the sphere radius;

[0092] The general equation for fitting a sphere by the RANSAC spherical fitting algorithm is:

[0093] ;

[0094] In the formula, D , E , F , G are fitting coefficients, is the sphere center coordinate, and the sphere radius .

[0095] The RANSAC spherical fitting algorithm is iteratively verified to obtain the optimal fitting solution, i.e. the optimal D , E , F , G At this time, the sphere obtained by fitting has strong noise resistance.

[0096] More specifically, before performing sphere fitting, the parameters of the RANSAC spherical fitting algorithm are set, and are adjusted according to the characteristics of the round steel pipe member and the scanning accuracy:

[0097] ​① Iterative random sampling point number is 4 points, and the number of iterations is 50000 times. Through the point cloud data, the noise points (such as scanning interference points) may exist in the spherical fitting. If the number of iterations is too small, it is easy to cause the fitting of the wrong sphere due to the sampling of noise points, and the number of iterations is too large, which will cause too much calculation. The iteration amount of 50000 times can greatly improve the probability of "sampling to 4 pure spherical nodes (without noise)", and ensure that the final fitted sphere center and radius are highly consistent with the true spherical node.

[0098] ② Distance threshold: set the maximum allowed distance of the point to the straight line (inner point judgment standard), which can be taken as 0.002. Since the slight radius error of the rod exists usually ≤1mm, the three-dimensional scanning error such as the precision of laser scanner is about 1~2mm, the value is added to the scanning accuracy (such as LiDAR equipment error) and the requirement of structure modeling, which can eliminate noise interference and retain the real details of the spherical node.

[0099] ③ Radius effective range: the value is [0.045, 0.18]. The radius of the spherical node of the spatial rod combination structure in engineering is usually in a small interval, and the range is determined by the measured point cloud data and engineering experience. The value ensures that the fitted sphere is a real spherical node, rather than a false sphere formed by noise or rod interference.

[0100] In the process of spherical fitting, 4 points P1, P2, P3 and P4 not in the same plane are selected in the spherical node candidate cluster in each iteration, and their coordinates are represented as , , , The coordinates of P1, P2, P3 and P4 are substituted into the spherical equation to obtain D , E , F , G , and then the sphere center coordinates and the sphere radius are calculated.

[0101] For each spherical node candidate cluster, the center point of the spherical node candidate cluster is taken as the initial reference position of the spherical fitting , and according to the number of iterations, 5000 times of spherical fitting are carried out.

[0102] Substep 1043: According to the radius effective range and the distance deviation of all points in the spherical node candidate cluster to the sphere, the spherical fitting result is screened, and the geometric center coordinates of the spherical node are determined.

[0103] Firstly, the number of iterations is 50000 times, and the validity of the fitted data is screened. If , the fitting result is directly discarded.

[0104] Then, the radiusr The fitting results in the range of [0.045, 0.18] are subjected to optimality screening: for each fitted sphere, the distance deviation of all points in the sphere node candidate cluster to the sphere is calculated, denoted as:

[0105] ;

[0106] In the formula, represents the Euclidean distance between any point in the sphere node candidate cluster and the final sphere center obtained by fitting, is the final sphere center coordinate obtained by fitting.

[0107] If , the point is an inner point of the sphere, and the number of inner points is counted. As the iteration proceeds, the number of inner points calculated each time is compared with the previous optimal result, and the sphere parameters corresponding to the fitting result with the most inner points are retained as the current optimal sphere parameters. After all 50000 iterations are completed, the sphere with the most inner points and the smallest residual is finally selected as the optimal fitting result, and a least squares refinement calculation is performed on the corresponding inner point set to obtain the final sphere center coordinate and sphere radius, realizing high-precision geometric recognition of the sphere node.

[0108] For all sphere node candidate clusters, the sphere node candidate clusters whose number of inner points accounts for ≥80% of the total number of points in the sphere node candidate cluster are regarded as "high-density effective clusters", and the sphere node candidate clusters whose number of inner points accounts for <80% of the total number of points in the sphere node candidate cluster are removed.

[0109] Finally, the sphere center coordinates of the sphere node corresponding to each high-density effective cluster are taken as the geometric center coordinates of each sphere node.

[0110] Step 2: based on point cloud matching degree filtering and sphere intersection filtering, determining the spatial axis of the rod between the sphere nodes, determining the spatial position and direction of the rod, and constructing a two-dimensional model of the rod;

[0111] Optionally, referring to FIG. 8, step 2 specifically includes the following substeps: Figure 2

[0112] Substep 201: based on the geometric center coordinates of the sphere nodes, calculating the distance between any two sphere nodes, screening effective node pairs according to the distance between the sphere nodes, and generating an initial rod axis candidate set;

[0113] Selecting the geometric center coordinates , of any two sphere nodes as the sphere center coordinates of the sphere nodes, and calculating the straight-line distance between the sphere centers of the two sphere nodes by using a three-dimensional Euclidean distance formula, the calculation formula being:​

[0114] ;

[0115] In the formula, The center of the sphere represents the two sphere nodes. and The straight-line distance between them (hereinafter referred to as the line segment length) reflects the spatial length of the potential member;

[0116] Based on the typical length constraints of the members in the space frame structure of the project, the distances that satisfy the constraints are selected. (Node pairs with units consistent with point cloud coordinates) This refers to valid node pairs. Each valid node pair corresponds to a candidate initial member axis. All candidate initial member axes constitute the candidate set of initial member axes, and each candidate initial member axis (is a line segment, denoted as line segment) is further defined as a line segment. The two endpoints of the sphere are the centers of the two sphere nodes. and The spatial connection relationship of the rods was initially determined.

[0117] Sub-step 202: Based on point cloud matching degree filtering, select first-level candidate members from the initial candidate set of member axes;

[0118] For any line segment First calculate the line segment Coordinates of the midpoint M .

[0119] Based on the KDTree index of the pole point cloud, pole points within a two-unit radius around the midpoint M of the candidate initial pole axis are extracted to form a pole point set. S , This ensures that only valid member points near the line segment are analyzed; where, These represent the points of the member within a 2-unit radius around the midpoint M of the candidate initial member axis. and , Indicates the number of points on the link.

[0120] For each extracted rod point Calculate its distance to the line segment The distance is calculated using vector projection. Line segment direction vector Projection scale t and to t Trim it to constrain it within the range [0,1] to obtain (If the range is exceeded, select the endpoints) to calculate the line segment. upper distance The coordinates of the latest point N are represented as:

[0121] , and the distance of point to line segment is:

[0122] ;

[0123] In the formula, is the distance of point to line segment , is the coordinates of point , and is the coordinates of point N.

[0124] Suppose the distance threshold of the rod point to line segment is 0.20 (adapted to the point cloud accuracy and the pipe radius), and the number of rod points in the rod point set S that satisfy is counted.

[0125] In order to adapt to the point cloud density requirements of rod pieces of different lengths, the dynamic filtering threshold is set to (that is, 5 times the length of the line segment), which can avoid the deletion of rod pieces due to insufficient point cloud. If the number of rod points in the rod point set S that satisfy is greater than the dynamic filtering threshold, it indicates that the candidate initial rod axis has sufficient rod point cloud data support, and is retained as a primary candidate rod, otherwise it is deleted, and the first stage of filtering is completed.

[0126] Substep 203: Based on the sphere intersection filtering, the rod pieces intersecting with three or more than three sphere nodes are removed from the primary candidate rod pieces, and the effective rod pieces connecting only two sphere nodes are retained.

[0127] After the primary candidate rod pieces are filtered based on the point cloud matching degree, there may still be invalid cases of crossing redundant sphere nodes (such as line segments crossing non-connected sphere nodes obliquely), which need to be further screened through the intersection judgment of line segments and sphere nodes to ensure that the rod pieces only connect two sphere nodes.

[0128] Each sphere node corresponds to a virtual sphere, and the sphere center of the sphere node represents the virtual sphere, and the radius threshold of the virtual sphere is set. For the primary candidate rod piece , it is judged whether it intersects with the sphere center . The core is to calculate the shortest distance from the sphere center to the line segment , and the calculation method is consistent with the calculation method of the distance of point to line segment . If Then the line segment With the center of the ball Intersection indicates that line segments Pass through the center of the ball ;like Then line segment With the center of the ball They do not intersect.

[0129] Statistics and Line Segments Number of intersections of intersecting sphere nodes s ,like This indicates that the line segment Passing through 3 or more ball joints does not conform to the engineering logic of "a member only connects two joints" and is therefore rejected; if This indicates that the line segment Only with two ball nodes (i.e., endpoints) and The intersections are consistent with the actual engineering requirements and are retained as valid members.

[0130] Sub-step 204: Based on the spatial axis of the effective members, determine the spatial position and orientation of the members, and construct a two-dimensional model of the members;

[0131] The effective member is taken as the spatial axis of the member, and the two endpoints of the member correspond to the centers of the two sphere nodes. , Precisely determine the connection positions at both ends of the rod; line segment The direction is determined by the direction vector. By defining the members, we can finally obtain a two-dimensional model of the rod.

[0132] Step 3: Extract the engineering drawing information of the spatial pole system combination structure of the existing building, transform the coordinate system of the two-dimensional model of the pole, and determine the ball node information and pole information after the coordinate system transformation;

[0133] Optionally, refer to the appendix Figure 3 As shown, step 3 includes the following sub-steps:

[0134] Sub-step 301: Extract the engineering drawing information of the spatial pole system combination structure of the existing building, and transform the local coordinate system of the two-dimensional model of the pole to be consistent with the global coordinate system of the engineering drawing based on the ball node information;

[0135] Extract engineering drawing information of the spatial frame structure of existing buildings, and obtain the global coordinate system of the engineering drawings from the title block or instruction manual. ,in, The origin of the global coordinate system (such as building corners, axis intersections, etc.); X The axis refers to the "longitudinal axis direction" (such as the building length direction) marked on the engineering drawings.Y The axis is the "transverse axis direction" (such as the building width direction) marked on the engineering drawing; the coordinate unit is usually mm or m (unified with the model unit, preferably m). Then the core data of the engineering drawing is directly exported or manually measured (the precision needs to be ≤0.1 mm) using CAD software (such as AutoCAD). In this step, the feature nodes in the engineering drawing need to be one-to-one corresponding to the feature nodes in the two-dimensional model. In the present application, a spherical node is selected as the feature node.

[0136] The coordinate system of the two-dimensional model is defined as a local coordinate system The local coordinate system is usually not coincident with the global coordinate system of the engineering drawing, so the local coordinate system of the two-dimensional model needs to be transformed to be consistent with the global coordinate system of the engineering drawing through coordinate transformation, to ensure that the coordinates of the same spherical node are consistent in the two systems.

[0137] The origin of the local coordinate system is set The coordinate in the global coordinate system is a translation vector ; the angle between the axis of the local coordinate system and the axis of the global coordinate system is the rotation angle (counterclockwise is positive and clockwise is negative); the local coordinate of any spherical node in the two-dimensional model is , and the coordinate in the global coordinate system after transformation is . The local coordinate system is rotated and translated to obtain the global coordinate system by using the trigonometric relationship, and the transformation parameters are determined by matching the coordinates of the "spherical nodes of the two-dimensional model" and the "spherical nodes of the engineering drawing". Due to the existence of measurement errors, the least squares method is used for optimization to ensure the minimum error. The transformation parameters obtained by the least squares method are:

[0138] ;

[0139] In the formula, , represents the average of the axis coordinates of the X spherical nodes in the engineering drawing, h represents the count of the feature nodes; , represents the average of the H axis coordinates of the spherical nodes in the two-dimensional model, and so on.

[0140] ;

[0141] The above formula can be solved to obtain In addition, the direction of the coordinate system needs to be verified to ensure that In a reasonable range, such as .

[0142] Obtain the transformation parameters After that, the local coordinate system is rotated around the origin by an angle , and the coordinates of the spherical nodes are temporarily transformed into temporary coordinates :

[0143] ;

[0144] That is, the rotation matrix of the local coordinate system is , which essentially decomposes into the direction of the global coordinate system .

[0145] Translate the temporary coordinates obtained after rotation along to obtain the coordinates of the spherical nodes in the global coordinate system :

[0146] ;

[0147] The spherical nodes in the transformed two-dimensional model are one-to-one matched with the spherical nodes in the engineering drawing.

[0148] Sub-step 302: Based on the transformed local coordinate system, the bar in the two-dimensional model is preliminarily positioned, and then the deviation calculation and correction of the bar in the two-dimensional model are performed to determine the bar information after the coordinate system transformation;

[0149] Specifically, the deviation of the length of the bar in the two-dimensional model and the length of the bar in the engineering drawing is defined as:

[0150] ;

[0151] wherein is the length of the bar in the engineering drawing, is the length of the bar in the two-dimensional model.

[0152] The threshold value of the allowable deviation of the length of the bar in the spatial truss combination structure is set to 1 mm , if , it is qualified and does not need to be corrected, if , the coordinates of the feature nodes in the two-dimensional model need to be corrected so that the length of the corrected bar satisfies .

[0153] ​The midpoint deviation of the midpoint of the axis of the rod in the two-dimensional model and the global coordinates of the midpoint of the axis of the rod in the engineering drawing are compared :

[0154] ;

[0155] In the formula, is the global coordinates of the midpoint of the axis of the rod in the two-dimensional model, is the global coordinates of the midpoint of the axis of the rod in the engineering drawing.

[0156] The midpoint deviation threshold of the midpoint of the axis of the rod in the spatial rod system combination structure is set to 0.3 mm , if , then it is qualified and does not need to be corrected, otherwise the coordinate transformation needs to be rechecked or the coordinates of the feature nodes in the two-dimensional model need to be corrected.

[0157] Step 4: Based on the spherical node information and the rod information after the coordinate system transformation, a three-dimensional spherical node model and a three-dimensional rod model are generated, and BIM attribute mapping is performed on the generated three-dimensional rod model to complete BIM modeling.

[0158] Optionally, step 4 includes the following sub-steps,

[0159] Sub-step 401: Establish a BIM global coordinate system, and generate a three-dimensional spherical node model based on the spherical node information after the coordinate system transformation;

[0160] The data of the two-dimensional model is converted into structured data that is mechanically readable, and a BIM global coordinate system is established, which is consistent with the global coordinate system of the engineering drawing. For any spherical node , the three-dimensional coordinates of which are , wherein is the coordinate position obtained according to the coordinate transformation and correction that meets the accuracy requirement, is a preset elevation.

[0161] In the Python program, the nodes array in the structured data is read through the “Data.ImportJSON” node, the of each spherical node is extracted, is assigned; then the “Point.ByCoordinates (X, Y, Z)” node is called to generate a three-dimensional node; finally, the “FamilyInstance.ByPoint” node is called to associate the three-dimensional node with the “net rack node family” (such as a circular node plate family) in BIM, and the three-dimensional spherical node model is generated.

[0162] Sub-step 402: According to the ball node position and the coordinate-transformed bar information, the bar axis and the spatial pose of the bar section are determined, and a three-dimensional bar model is generated;

[0163] The three-dimensional bar model generation needs to determine the bar axis (line segment and length) and the spatial pose (rotation + translation) of the bar section. Therefore, the three-dimensional bar model generation can further include the following sub-steps:

[0164] Sub-step 4021: The start point coordinates and the end point coordinates of the bar are matched from the structured data, and the length of the bar is determined;

[0165] The start point P s of the bar is matched from the structured data, and the coordinates of the start point P s are represented as , net represents the elevation, the end point P e of the bar is matched from the structured data, and the coordinates of the end point P e are represented as ; According to the start point of the bar and the end point of the bar, the line segment parameter equation of the bar is determined as:

[0166] ;

[0167] In the formula, represents the line segment vector of the bar, represents the direction vector of the bar, , is the direction coefficient, corresponds to , corresponds to .

[0168] The length of the bar axis is verified by the two-point distance formula, and the final result calculated is the length of the bar.

[0169] Sub-step 4022: The spatial pose of the bar section is determined, and the generation of the three-dimensional bar model is completed;

[0170] Since the axis of the bar in the two-dimensional model is in the XY plane ( ), the direction vector is , therefore, the section of the bar needs to be rotated around the axis to ensure that its normal vector is consistent with the direction of the axis (i.e. the section plane is perpendicular to the axis in the XY plane), so as to finally convert the two-dimensional contour into a three-dimensional section entity.

[0171] First, take any point (such as the midpoint) on the axis of the bar as the center P center of the section, and the coordinates of the center P center are , then the direction vector is obtained by trigonometric function and X Angle between axes , which is the angle by which the cross-section of the member needs to be rotated about the axis, and the specific calculation formula is:

[0172] ;

[0173] In the formula, Direction vectors Projected to X shaft and Y The length on the axis.

[0174] when along X Positive axis direction hour, The cross-section does not need to be rotated; when along Y Positive axis direction At that time, the cross section needs to be around Axis rotation .

[0175] Therefore, the rotation matrix of the cross section can be obtained as follows: The coordinates of all vertices of the two-dimensional cross-sectional profile of the rod. Multiply The coordinates of the vertices of the rotated 3D cross section are obtained. Ensure that the interface is perpendicular to the axis of the rod.

[0176] Based on the determination of the spatial orientation of the member cross-sections, the Python program reads the `members` array from the structured data, iterates through each member, and then matches the starting point P of the member. s and endpoint P e , call Generate the axis of the member, then read the section parameters corresponding to section_id, call FamilySymbol.LookupByName(doc, "Pipe D100x6") to get the section family type; call StructuralFraming.Create(doc, section_symbol, line) to generate the member entity based on the axis and section (Revit automatically handles section rotation and translation), and finally repeat the above steps until all 3D member models are generated.

[0177] Sub-step 403: Perform BIM attribute mapping on the generated 3D rod model to complete BIM modeling and verify accuracy.

[0178] With the "Element.SetParameterByName" node, select the generated three-dimensional rod; extract the attribute value in the structured data (such as "material name = Q355B"), and then match the corresponding parameters of the three-dimensional rod (such as "material"), and assign the attribute value to the parameter. Repeat the above steps to batch process all rods, complete the BIM attribute mapping. After completing this step, the geometric accuracy of the BIM model is verified by methods such as ball node position deviation calculation, rod length deviation calculation, and attribute integrity check, to ensure the accuracy of the BIM model and avoid errors. After the final verification, integrate the ball nodes and rods into the BIM model as a whole, and export the format (such as.rvt,.ifc) according to the engineering requirements, for subsequent structural analysis (such as ETABS link), construction simulation or engineering quantity statistics.

[0179] Simulation verification:

[0180] The simulation experiment uses laser point cloud to scan a certain existing building structure to obtain the collection results of point cloud data as shown in the accompanying Figure 4 The BIM modeling result finally established by the BIM modeling method in the application is shown in the accompanying Figure 5 Comparing the accompanying Figure 4 and the accompanying Figure 5 It can be seen that the BIM modeling method in the application can accurately capture the ball nodes and the spatial rod structure, and the finally generated BIM model can accurately reflect the three-dimensional geometric information of the existing building.

[0181] Example two:

[0182] Example two provides a kind of existing building spatial truss combined structure reverse BIM modeling system for realizing the reverse BIM modeling method of the spatial truss combined structure of the existing building described in example one.

[0183] Specifically, the system comprises:

[0184] The ball node positioning module is used to obtain the point cloud data of the spatial truss combined structure of the existing building, and determine the geometric center coordinates of the ball nodes.

[0185] The rod positioning module is based on point cloud matching degree filtering and sphere intersection filtering to determine the spatial axis of the rod between the ball nodes, determine the spatial position and direction of the rod, and construct the two-dimensional model of the rod.

[0186] The coordinate transformation module is used to extract the engineering drawing information of the spatial truss combined structure of the existing building, transform the coordinate system of the two-dimensional model of the rod, and determine the ball node information and rod information after coordinate system transformation.

[0187] The BIM modeling module generates 3D sphere node models and 3D rod models based on the sphere node information and rod information after coordinate system transformation, and performs BIM attribute mapping on the generated 3D rod models to complete BIM modeling.

[0188] Example 3:

[0189] Embodiment 3 provides an electronic device, including at least one processor; and a memory communicatively connected to the processor; wherein the memory stores instructions that are executed by the processor, the instructions being executed by the processor to enable the processor to perform the reverse BIM modeling method for existing building space frame combination structures described in Embodiment 1.

[0190] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for reverse BIM modeling of existing building spatial frame composite structures, characterized by: The method includes the following steps: Obtain point cloud data of the spatial pole system combination structure of the existing building and determine the geometric center coordinates of the spherical nodes; Based on point cloud matching degree filtering and sphere intersection filtering, the spatial axis of the rod between sphere nodes is determined, the spatial position and orientation of the rod are determined, and a two-dimensional model of the rod is constructed. Extract engineering drawing information of the spatial pole system combination structure of the existing building, transform the coordinate system of the two-dimensional model of the pole, and determine the ball node information and pole information after the coordinate system transformation; Based on the ball node information and rod information after coordinate system transformation, a three-dimensional ball node model and a three-dimensional rod model are generated, and BIM attribute mapping is performed on the generated three-dimensional rod model to complete BIM modeling. The process of determining the spatial axis of the rods between sphere nodes, the spatial position and orientation of the rods, and constructing a two-dimensional model of the rods based on point cloud matching degree filtering and sphere intersection filtering includes the following steps: Based on the geometric center coordinates of the ball nodes, calculate the distance between any two ball nodes, filter out valid node pairs based on the distance between the ball nodes, and generate an initial candidate set of member axes; Based on point cloud matching degree filtering, first-level candidate members are selected from the initial candidate set of member axes; Based on sphere intersection filtering, members that intersect with three or more sphere nodes are removed from the first-level candidate members, and valid members that only connect two sphere nodes are retained; Based on the spatial axis of the effective members, determine the spatial position and orientation of the members, and construct a two-dimensional model of the members.

2. The reverse BIM modeling method for existing building spatial frame combination structures according to claim 1, characterized in that: The process of acquiring point cloud data of the spatial pole assembly structure of an existing building and determining the geometric center coordinates of the spherical nodes includes the following steps: The spatial frame structure of the existing building is scanned using a laser scanner to obtain point cloud data of the spatial frame structure. Denoising preprocessing is performed on the point cloud data of the spatial pole system assembly structure; Points with curvature greater than a threshold are selected from the preprocessed point cloud data and used as candidate sphere nodes. Based on the DBSCAN clustering method and the RANSAC spherical fitting algorithm, the geometric center coordinates of the sphere nodes are determined by fitting a sphere to the candidate points of the sphere nodes.

3. The reverse BIM modeling method for existing building spatial frame composite structures according to claim 2, characterized in that: The method based on DBSCAN clustering and RANSAC spherical fitting algorithm, which determines the geometric center coordinates of sphere nodes by fitting a sphere to candidate sphere nodes, includes the following steps: Perform DBSCAN clustering on all candidate ball nodes and discretize them into multiple candidate ball node clusters, and determine the center point of each candidate ball node cluster; For each candidate cluster of sphere nodes, the RANSAC sphere fitting algorithm is used to fit the sphere, and the coordinates of the sphere center and the radius of the sphere are calculated. Based on the effective radius range and the distance deviation from the sphere to the sphere of all points in the candidate cluster of sphere nodes, the sphere fitting results are screened, and the geometric center coordinates of the sphere nodes are determined.

4. The reverse BIM modeling method for existing building spatial frame combination structures according to claim 3, characterized in that: The point cloud matching degree-based filtering for selecting first-level candidate members from the initial candidate set of member axes specifically includes the following steps: For any candidate initial member axis, calculate the coordinates of the midpoint of the candidate initial member axis; Based on the KDTree index of the rod point cloud, the rod points within a two-unit radius around the midpoint of the candidate initial rod axis are extracted to form a rod point set; For each member point in the member point set, calculate the member point. S n Distance to the axis of the candidate initial member ; Set a distance threshold and count the number of pole points in the set. The number of pole points less than or equal to the distance threshold; Set the dynamic filtering threshold to If the set of member points satisfies If the number of pole points less than or equal to the distance threshold is greater than the dynamic screening threshold, they are retained as first-level candidate poles; otherwise, they are discarded. is the line segment length of the candidate initial rod axis.

5. The reverse BIM modeling method for existing building spatial frame combination structures according to claim 4, characterized in that: The process of extracting engineering drawing information of the spatial frame structure of existing buildings, transforming the coordinate system of the two-dimensional model of the frame members, and determining the spherical node information and frame member information after the coordinate system transformation includes the following steps: Extract the engineering drawing information of the spatial pole system combination structure of the existing building, and transform the local coordinate system of the two-dimensional model of the pole to be consistent with the global coordinate system of the engineering drawing based on the ball node information, so as to obtain the ball node information after coordinate system transformation. Based on the transformed local coordinate system, the members in the two-dimensional model are initially located. Then, the deviation of the members in the two-dimensional model is calculated and corrected to determine the member information after the coordinate system transformation.

6. The reverse BIM modeling method for existing building spatial frame composite structures according to claim 5, characterized in that: The process of generating 3D spherical node models and 3D rod models based on the coordinate system-transformed spherical node information and rod information, and performing BIM attribute mapping on the generated 3D rod models to complete BIM modeling includes the following steps: Establish a global BIM coordinate system and generate a 3D spherical node model based on the spherical node information after coordinate system transformation. Based on the ball node position and the rod information after coordinate system transformation, determine the spatial orientation of the rod axis and the rod cross section, and generate a three-dimensional rod model; Perform BIM attribute mapping on the generated 3D rod model to complete BIM modeling, and verify the accuracy.

7. A reverse BIM modeling system for existing building spatial frame structures, used to implement the reverse BIM modeling method for existing building spatial frame structures as described in any one of claims 1-6, characterized in that: The system includes: The sphere node positioning module is used to acquire point cloud data of the spatial pole system combination structure of existing buildings and determine the geometric center coordinates of the sphere nodes. The pole positioning module, based on point cloud matching degree filtering and sphere intersection filtering, determines the spatial axis of the pole between sphere nodes, determines the spatial position and orientation of the pole, and constructs a two-dimensional model of the pole. The coordinate transformation module is used to extract engineering drawing information of the spatial rod system combination structure of existing buildings, transform the coordinate system of the two-dimensional model of the rod, and determine the ball node information and rod information after the coordinate system transformation. The BIM modeling module generates 3D sphere node models and 3D rod models based on the sphere node information and rod information after coordinate system transformation, and performs BIM attribute mapping on the generated 3D rod models to complete BIM modeling.

8. An electronic device, characterized in that, The device includes at least one processor; and a memory communicatively connected to the processor; wherein the memory stores instructions that are executed by the processor to enable the processor to perform the reverse BIM modeling method for existing building space frame structure as described in any one of claims 1-6.

Images