BIM-based curtain wall simulation design method and system

Abstract

The invention relates to the technical field of building model design, and discloses a BIM-based curtain wall simulation design method and system, and the method comprises the steps: obtaining building point cloud data containing complex curved surface features; combining various geometric features of the point cloud data to improve a traditional region growing algorithm to segment the point cloud data; and fitting geometric primitives with the segmented point cloud data, and constructing a building BIM model to complete curtain wall simulation design. The invention aims to improve the accuracy of edge segmentation of point cloud data of different building planes by a region growing algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of architectural model design, in particular to a curtain wall simulation design method and system based on BIM. BACKGROUND

[0002] With the development of the construction industry, building curtain walls, as an important part of the building facade, have multiple functions such as heat preservation, heat insulation, sound insulation, and aesthetics. Building information modeling (BIM) is a digital modeling technology that integrates various stages such as architectural design, construction, and operation management. BIM creates a three-dimensional digital model by collecting information data of buildings, integrates and collaborates various information in architectural design, and realizes the visualization and digitization of the architectural engineering simulation design process. The curtain wall simulation design method based on BIM realizes the full-process simulation and optimization of building curtain wall design, improving the collaborative efficiency of design and construction.

[0003] However, in modern architectural design, in order to highlight the novel design concept and artistic effect, many buildings are designed with complex curved surface design style. When collecting the appearance data (such as point cloud data) of these buildings to construct the BIM model, due to the massive and complex geometric characteristics, the existing BIM model construction method has low accuracy in extracting geometric features from the point cloud data of the building and segmenting the point cloud data, resulting in a deviation between the generated BIM model and the actual building. This leads to problems and conflicts in the actual construction of the curtain wall simulation design scheme based on the BIM model, affecting the construction efficiency and reducing the construction quality. SUMMARY

[0004] To solve the above technical problems, the present application provides a curtain wall simulation design method and system based on BIM, and the technical solutions adopted are as follows: In a first aspect, an embodiment of the present application provides a curtain wall simulation design method based on BIM, which comprises the following steps: Obtain building point cloud data containing complex curved surface features; Voxelize the point cloud data, and cluster the point cloud voxels based on the density features and curvature features of the point cloud voxels to select region growth seed points; Construct an adaptive growth criterion based on the curvature difference and normal vector difference between any to-be-merged point cloud voxel and the region of any seed point, and adjust the merging threshold in the adaptive growth criterion based on the fitting surface error of the to-be-merged point cloud voxel and the region of the seed point to perform region growth on the point cloud voxels; Fit the segmented point cloud data to geometric primitives and construct a building BIM model to complete the curtain wall simulation design.

[0005] Preferably, the distance metric in the clustering of the point cloud voxels is positively correlated with the density feature difference and the curvature feature difference between the point cloud voxels, respectively.

[0006] Preferably, the selection of the region growing seed point comprises: selecting three point cloud voxels between the maximum distance and the minimum distance from the distances of all point cloud voxels in each cluster to the cluster center point cloud voxel as the seed points for region growing, and the cluster center point cloud voxel.

[0007] Preferably, the adaptive growth criterion comprises: When the number of point cloud voxels in the region where the seed point is located is greater than or equal to a preset number, an adaptive threshold weight is constructed based on the fitting surface error of the point cloud voxel to be merged and the region where the seed point is located; otherwise, the adaptive threshold weight is assigned as 1; The adaptive threshold weight is used as the weight of the preset curvature feature difference threshold and the preset normal vector feature difference threshold, respectively. When the curvature difference is less than or equal to the weighted preset curvature feature difference threshold, and the normal vector difference is less than or equal to the weighted preset normal vector difference threshold, it is determined that the point cloud voxel to be merged is a merging condition of the region where the seed point is located, and the point cloud voxel to be merged is merged into the region; otherwise, no merging is performed.

[0008] Preferably, the curvature difference is the difference between the curvature feature of the point cloud voxel to be merged and the mean value of the curvature features of all point cloud voxels in the region where the seed point is located; and the curvature feature of the point cloud voxel is calculated by eigenvalue decomposition of the covariance matrix of the centroid point cloud in the point cloud voxel.

[0009] Preferably, the normal vector difference is the difference between the normal vector feature of the point cloud voxel to be merged and the mean value of the unit normal vector features of all point cloud voxels in the region where the seed point is located; and the normal vector feature of the point cloud voxel is determined by the eigenvector corresponding to the minimum eigenvalue of the covariance matrix of the centroid point cloud in the point cloud voxel.

[0010] Preferably, the method for calculating the fitting surface error of the point cloud voxel to be merged and the region where the seed point is located comprises: When the number of point cloud voxels in the region where the seed point is located is greater than or equal to a preset number, the spatial distance between the point cloud voxel to be merged and the seed point is obtained; Taking the seed point as the center, a circle is made by using the spatial distance to determine the radius, all point cloud voxels in the region where the seed point is located in the circle are fitted to a surface, the geometric distance from the point cloud voxel to be merged to the surface is calculated as the fitting residual of the point cloud voxel to be merged; The fitting residual mean of the to-be-merged point cloud voxel and its eight neighborhood voxel points and the surface is taken as the fitting surface error of the to-be-merged point cloud voxel and the region where the seed point is located.

[0011] Preferably, the adaptive threshold weight is constructed based on the fitting surface error of the to-be-merged point cloud voxel and the region where the seed point is located, comprising: , wherein is the adaptive threshold weight constructed based on the fitting surface error of the to-be-merged point cloud voxel and the region where the first seed point is located, e is a natural constant, is a preset value, is the adaptive threshold weight constructed based on the fitting surface error of the to-be-merged point cloud voxel and the region where the first seed point is located.

[0012] Preferably, the adaptive threshold weight is taken as the weight of the preset curvature feature difference threshold and the preset normal vector feature difference threshold, respectively, that is, the adaptive threshold weight is multiplied by the preset curvature feature difference threshold and the preset normal vector feature difference threshold, respectively.

[0013] In a second aspect, another embodiment of the present application also provides a BIM-based curtain wall simulation design system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the BIM-based curtain wall simulation design method described above when executing the computer program.

[0014] The present application has at least the following beneficial effects: The present application solves the problem that the traditional region growing segmentation algorithm has poor segmentation effect on point cloud data by randomly selecting a seed point and formulating a growth criterion based on the curvature feature of the point cloud. First, the density and curvature feature of the point cloud voxel are considered to cluster the point cloud voxel and select the seed point of the region growing, which can avoid the interference of noise points when selecting the seed point, better cover the point cloud voxel of different geometric feature regions, and improve the representativeness of the seed point. Then, the growth criterion is constructed according to the curvature feature difference and normal vector feature difference of different building plane point cloud data, and the adaptive weight of the growth criterion threshold is constructed according to the fitting surface error between each to-be-merged point cloud voxel and the region where the seed point is located, which improves the accuracy of the region growing algorithm in segmenting the edge of different building plane point cloud data, and facilitates the subsequent construction of a high-precision BIM model for simulation design of a curtain wall scheme. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0016] Figure 1 The flowchart of the BIM-based curtain wall simulation design method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0017] Embodiment 1 The BIM-based curtain wall simulation design method provided by an embodiment of the present application, with reference to Figure 1 The method comprises the following steps: Step 1: Obtain building point cloud data containing complex curved surface features.

[0018] The unmanned aerial vehicle is equipped with a laser radar device to obtain three-dimensional point cloud data of a building containing complex curved surface features.

[0019] The obtained building point cloud data is preprocessed, and a statistical outlier algorithm is used to remove noise points to improve data quality. The statistical outlier algorithm is prior art, and its specific process will not be repeated.

[0020] Step 2: Combine various geometric features of the point cloud data to improve the traditional region growing algorithm to segment the point cloud data.

[0021] Since the collected building point cloud data contains a large amount of complex curved surface building surface point cloud data, the point cloud data of these curved surface regions has more complex geometric structure features. The traditional method of constructing a BIM model based on point cloud data is difficult to accurately extract the geometric features of the curved surface point cloud when processing a large amount of complex curved surface point cloud data. For example, the region growing-based segmentation algorithm randomly selects seed points and formulates growth criteria based on point cloud curvature features when segmenting point clouds, resulting in poor segmentation effect of point cloud data. This makes the geometric primitive fitting of local building point cloud data inaccurate when constructing a BIM model subsequently, and the geometric feature data of the building cannot be accurately obtained. Finally, the curtain wall simulation design scheme based on the BIM model does not match the actual construction, affecting the efficiency of curtain wall scheme construction.

[0022] To solve the above problems, the density and curvature features of the point cloud voxel are considered to cluster the point cloud voxel to obtain the seed point of the region growing, then the growth criterion is constructed according to the curvature consistency and normal vector consistency constraint features of different building plane point cloud data, and the edge saliency score is constructed as the adaptive weight of the growth criterion threshold considering the plane fitting error between each to-be-merged point cloud voxel and the seed point, so as to improve the accuracy of the edge segmentation of the region growing algorithm for different building plane point cloud data, and obtain clearer segmentation edges when segmenting different curved surfaces.

[0023] Firstly, the point cloud data is voxelized to improve the efficiency of subsequent calculation of feature differences between point clouds. The point cloud voxelization is to divide a series of voxels in a three-dimensional space from discrete point clouds, each voxel contains a plurality of point clouds, and then the features of the centroid point cloud in each voxel are taken as the features of the voxel, which include but are not limited to spatial position, curvature, normal vector difference and density feature. Specifically, a voxel grid coordinate system is created in a three-dimensional space, and the resolution of each voxel is , The size can be set by the implementer according to the scene, and is not specially limited. According to the building point cloud data collected in the scene, The size is set to 0.1m.

[0024] In the traditional region growing-based point cloud segmentation method, the seed point is often selected in a random way, which lacks consideration of the geometric features of the point cloud, and it is difficult to ensure that the data features of the selected seed point represent the different curved surfaces in the entire building point cloud data, which may cause the selected seed point to fall on a noise point or a boundary point, and easily cause over-segmentation or incomplete segmentation in the subsequent growth result.

[0025] For the obtained point cloud voxel data, wherein represents the i-th point cloud voxel, and it is assumed that contains a total of point clouds.

[0026] The density and curvature features of the point cloud voxel are considered to cluster the point cloud voxel to obtain the seed point of the region growing. The density feature of the point cloud voxel is , is the edge length of the point cloud voxel . The density represents the local compactness of the point cloud, and the number of point clouds in all point cloud voxels will not be completely the same. On a plane, the point clouds are uniformly distributed after projection, while for curved surfaces or glass cavities, the distribution of the point clouds presents a certain sparseness due to the influence of the curved surface angle reflection. The high-density region often corresponds to a smooth building plane, and the low-density region may be a curved surface or a glass cavity region.​

[0027] The curvature reflects the bending degree of the local surface in the building point cloud, the area with smaller curvature corresponds to a plane or a gentle curved surface, and the area with larger curvature represents a corner and a complex curved surface. For the point cloud voxel , the covariance matrix of the centroid point cloud is calculated, and then the eigenvalues 、 、 of the covariance matrix are obtained by eigenvalue decomposition, and the curvature feature of the point cloud voxel is obtained. The covariance matrix is calculated by the neighborhood of the point cloud, and the calculation of the covariance matrix and the eigenvalue decomposition are prior art, and the specific process is not described again.

[0028] According to the above analysis, the distance measurement when the point cloud voxel is clustered is positively correlated with the density feature difference and the curvature feature difference between the point cloud voxels.

[0029] It can be understood that the positive correlation means that the dependent variable will increase with the increase of the independent variable, and the dependent variable will decrease with the decrease of the independent variable, which is determined by actual application, and the present application does not make special limitation.

[0030] In the embodiment, the distance measurement formula when the point cloud voxel is clustered is further represented as: wherein Sig is a Sigmoid function, which maps the density feature and the curvature feature of the point cloud voxel to [0, 1], 、 represent the density features of the first 、 point cloud voxel, respectively, 、 represent the curvature features of the first 、 point cloud voxel, respectively. The point cloud voxel is clustered by considering its density and curvature features, and finally all point cloud voxels are divided into point cloud voxel clusters with similar density and curvature features. The point cloud voxel clustering method adopts a K-means algorithm, The size can be set by the implementer according to the implementation scene, and no special limitation is made, and the volume of the modeled building is larger and the complexity of the building design is higher, which can appropriately increase the value of to find more point cloud voxels with different features. In the embodiment, The size is 100.

[0031] Further, select the distance between all point cloud voxels in each cluster to the cluster center point cloud voxel, the maximum distance to the minimum distance between the first, second, third quartile three point cloud voxels and the cluster center point cloud voxel as the seed point of region growing.

[0032] The seed point selected by this strategy can select representative point cloud voxels from each different feature cluster, avoiding the interference of noise points, better covering the point cloud voxels in the building point cloud data with different geometric feature regions, improving the representativeness and stability of the seed point, and facilitating more comprehensive region segmentation.

[0033] So far, the seed point selection strategy has been developed.

[0034] Further, the growth criterion of the region growing algorithm is constructed to segment different planes of point cloud data.

[0035] The traditional region growing algorithm judges the difference in normal vector and curvature feature based on a fixed threshold. This setting method of growth criterion will lead to inaccurate segmentation edges of different planes, and the reason for this phenomenon comes from the complexity and non-uniformity of building point cloud data itself. When collecting building point cloud data, it usually contains occlusion and glass mirror reflection effect, which will introduce measurement error and light reflection interference to cause point cloud scattering, so that the normal vector difference of voxels in the same plane region is larger than the actual difference. Using a fixed threshold will cause over-segmentation of plane edges, making the segmentation result inaccurate.

[0036] To solve this problem, the edge saliency score of the to-be-merged voxel is constructed as an adaptive weight of the threshold by considering the fitting surface residual feature of the to-be-merged voxel and the seed point region.

[0037] For the setting of the growth criterion, take the region of the first seed point and the to-be-merged point cloud voxel as an example to construct its growth criterion. Here, is one of the selected seed points.

[0038] First is the curvature difference: Where represents the curvature difference between the point cloud voxel and the region of the first seed point, represents the curvature feature of the point cloud voxel , and the calculation method is as described in the above steps, is the average curvature feature of all point cloud voxels in the region of the first seed point, and the calculation process is not repeated.

[0039] Then the normal vector difference is: wherein represents the normal vector feature of the point cloud voxel and the normal vector difference between the region where the th seed point is located, is the inverse cosine function, represents the normal vector feature of the point cloud voxel , that is, the eigenvector corresponding to the minimum eigenvalue of the covariance matrix of the centroid point cloud in the point cloud voxel , represents the mean value of the unit normal vector features of all point cloud voxels in the region where the th seed point is located, and is represented as follows: wherein represents the number of point cloud voxels in the region where the th seed point is located , represents the normal vector of the th point cloud voxel, represents the unitization of the normal vector, and ensures that the final is a normalized unit normal vector.

[0040] The determination feature of the growth criterion determined based on the curvature feature and the normal vector feature can comprehensively consider the normal vector difference and the curvature difference of the point cloud voxel as a constraint, and simultaneously depict the flatness and directionality of the local point cloud voxel, effectively distinguish the difference between the plane and various curved surfaces, and avoid affecting the segmentation accuracy by using a single feature.

[0041] Further, the surface equation of the region where the th seed point is located is fitted, when the number of point cloud voxels in the region is , for the point cloud voxel to be merged, the spatial distance between and the seed point is calculated by using the Euclidean distance. The Euclidean distance is a prior art, and its specific process will not be repeated. Then a circle is made with as the center and as the radius, and all point cloud voxels in the region within the circle are fitted to obtain a surface by using the least square surface fitting method. The least square surface fitting is a prior art, and its specific process will not be repeated. Then the fitting residual of the point cloud voxel is calculated That is, point cloud voxels To the surface Geometric distance: in express To the surface The geometric distance from a point to a surface is calculated using existing technology, and its specific process will not be elaborated further. A smaller residual indicates that the point better fits the fitted surface and has a higher probability of belonging to that region; conversely, a larger residual indicates that the point deviates from the fitted surface and may belong to another surface or be a boundary point between two surfaces.

[0042] Further calculations Eight neighboring voxels and surfaces The fitting residuals are obtained, including The mean of the fitting residuals of the nine point cloud voxels, including the one to be merged, is used as the point cloud voxels. With the The fitting surface error within the region where each seed point is located. . The larger it is, the more it indicates that... voxels and surfaces in the neighborhood of the center If the deviation is large, the voxel is more likely to be located at a region boundary or belong to a different surface. Therefore, the threshold will be lowered in the subsequent adaptive weighting to more strictly determine the value of that point; conversely... The smaller the value, the higher the threshold will be when used as an adaptive weight to avoid incorrect boundary segmentation caused by measurement errors and light reflection interference.

[0043] In summary, the growth criterion function, improved based on the fitting surface error within the region where the voxels of the point cloud to be merged and the seed point are located, can be expressed as: in Based on the voxels of the point cloud to be merged With the The adaptive threshold weights are constructed from the fitting surface error within the region of each seed point. where e is the natural constant. As a preset value, This is used to adjust the sensitivity of the adaptive threshold weight to the fitting surface error. Its value can be set by the implementer according to the implementation scenario, without special restrictions. In this embodiment... The size is 2. Specifically, when the region... In hour, If the merged area is too small, the threshold size will not be adjusted when the number of merged voxels is too small, taking into account the actual size of the building. is a preset curvature feature difference threshold, the size of which is taken as the curvature variance of all point cloud voxels, is a preset normal vector feature difference threshold, The size thereof can be set by the implementer according to the implementation scenario, without special limitation. In the embodiment, the size of the normal vector feature difference threshold is taken as the curvature variance of all point cloud voxels. .

[0044] When the curvature difference between the point cloud voxel and the fitting surface of the region to which the point cloud voxel belongs is less than the preset curvature feature difference threshold, the point cloud voxel satisfies the merging condition of the region where the first seed point is located, and the point cloud voxel is merged into the region ; otherwise, when the curvature difference is greater than the preset curvature feature difference threshold, the point cloud voxel is not merged.

[0045] The growth criterion thus designed can adaptively adjust the threshold size according to the fitting surface residual of the neighborhood of the point cloud voxel to be merged and the merged region, avoid the problem of inaccurate planar edge segmentation caused by the measurement error and light reflection interference introduced by the large amount of occlusion and glass mirror reflection effect in the point cloud data of the building, and improve the accuracy of subsequent modeling of the BIM model.

[0046] The above steps are repeated to complete the segmentation of the point cloud data.

[0047] Step three: fitting the segmented point cloud data to geometric primitives and constructing a building BIM model to complete the curtain wall simulation design.

[0048] The RANSAC algorithm is used to fit the geometric primitives of the segmented point cloud clusters to the segmented point cloud data, to obtain the geometric information of each point cloud cluster. The RANSAC algorithm is a prior art, and its specific process is not described again.

[0049] Finally, Revit or other software is used to automatically generate a parameterized building BIM model based on the segmented point cloud data and the fitted geometric primitive information, for completing the simulation design of the curtain wall scheme.

[0050] Embodiment 2 Another embodiment of the present application also provides a BIM-based curtain wall simulation design system, which comprises a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the BIM-based curtain wall simulation design method described above when executing the computer program.

[0051] ​​​​​​​Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.

[0052] It should be understood that the application is not limited to the precise construction and compositions described above and shown in the attached drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application.

Claims

1. A BIM-based curtain wall simulation design method, characterized in that, The method comprises the following steps: Obtain building point cloud data containing complex curved surface features; Voxelize the point cloud data, cluster the point cloud voxels based on the density features and curvature features of the point cloud voxels, and select a region growth seed point; Construct an adaptive growth criterion using the curvature difference and normal vector difference between any to-be-merged point cloud voxel and the region of any seed point, and adjust the merging threshold in the adaptive growth criterion using the fitting surface error of the to-be-merged point cloud voxel and the region of the seed point to perform region growth on the point cloud voxels; Fit the segmented point cloud data to geometric primitives and construct a building BIM model to complete the curtain wall simulation design.

2. The BIM-based curtain wall mockup design method of claim 1, wherein, The distance measurement when clustering the point cloud voxels is positively correlated with the density feature difference and the curvature feature difference between the point cloud voxels. 3.The BIM-based curtain wall simulation design method of claim 2, wherein, The selection of the region growth seed point comprises: selecting three point cloud voxels between the maximum distance and the minimum distance from the distances of all point cloud voxels in each cluster to the cluster center point cloud voxel as the seed points for region growth.

4. The BIM-based curtain wall mockup design method of claim 1, wherein, The adaptive growth criterion comprises: When the number of point cloud voxels in the region of the seed point is greater than or equal to a preset number, construct an adaptive threshold weight based on the fitting surface error of the to-be-merged point cloud voxel and the region of the seed point; otherwise, assign the adaptive threshold weight a value of 1; The adaptive threshold weight is used as the weight of the preset curvature feature difference threshold and the preset normal vector feature difference threshold, respectively; When the curvature difference is less than or equal to the weighted preset curvature feature difference threshold, and the normal vector difference is less than or equal to the weighted preset normal vector difference threshold, determine that the to-be-merged point cloud voxel meets the merging condition of the region of the seed point, and merge the to-be-merged point cloud voxel into the region; otherwise, do not perform merging.

5. The BIM-based curtain wall mockup design method of claim 4, wherein, The curvature difference is the difference between the curvature feature of the to-be-merged point cloud voxel and the average of the curvature features of all point cloud voxels in the region of the seed point; the curvature feature of the point cloud voxel is calculated by eigenvalue decomposition of the covariance matrix of the centroid point cloud in the point cloud voxel.

6. The BIM-based curtain wall mockup design method of claim 5, wherein, The normal vector difference is the difference between the normal vector feature of the to-be-merged point cloud voxel and the average of the unit normal vector features of all point cloud voxels in the region of the seed point; the normal vector feature of the point cloud voxel is determined by the eigenvector corresponding to the minimum eigenvalue of the covariance matrix of the centroid point cloud in the point cloud voxel.

7. The BIM-based curtain wall mockup design method of claim 4, wherein, The method for calculating the fitting surface error of the to-be-merged point cloud voxel and the region of the seed point comprises: When the number of point cloud voxels in the region of the seed point is greater than or equal to a preset number, obtain the spatial distance between the to-be-merged point cloud voxel and the seed point; Take the seed point as the center, determine the radius using the spatial distance, make a circle, fit the surface of all point cloud voxels in the region of the seed point in the circle, calculate the geometric distance of the to-be-merged point cloud voxel to the surface as the fitting residual error of the to-be-merged point cloud voxel; Take the average of the fitting residual errors of the to-be-merged point cloud voxel and its eight neighboring voxels and the surface as the fitting surface error of the to-be-merged point cloud voxel and the region of the seed point.

8. The BIM-based curtain wall mockup design method of claim 4 or 7, wherein, The adaptive threshold weight is constructed based on a fitting surface error of a voxel of a point cloud to be merged and a region where a seed point is located, and the adaptive threshold weight comprises: wherein is a fitting surface error of a voxel of a point cloud to be merged and a region where a seed point is located, is a fitting surface error of a voxel of a point cloud to be merged and a region where a seed point is located, is a fitting surface error of a voxel of a point cloud to be merged and a region where a seed point is located, is a natural constant, is a preset value, is a fitting surface error of a voxel of a point cloud to be merged and a region where a seed point is located, is a fitting surface error of a voxel of a point cloud to be merged and a region where a seed point is located.

9. The BIM-based curtain wall mockup design method of claim 8, wherein, The adaptive threshold weight is multiplied by the preset curvature feature difference threshold and the preset normal vector feature difference threshold respectively as the weight of the preset curvature feature difference threshold and the preset normal vector feature difference threshold. 10.A BIM-based curtain wall simulation design system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor implements the BIM-based curtain wall simulation design method in claim 1-7 when executing the computer program.

Images