CN-122023681-A - Civil engineering structure modeling method and system based on point cloud and electronic equipment

CN122023681ACN 122023681 ACN122023681 ACN 122023681ACN-122023681-A

Abstract

The embodiment of the specification discloses a modeling method, a system and electronic equipment of a civil engineering structure based on point clouds, wherein the modeling method comprises the steps of obtaining real scanning point cloud data and a design drawing of the civil engineering structure, carrying out image segmentation on the real scanning point cloud data to obtain independent point cloud data corresponding to each independent component, obtaining component modeling parameters based on the independent point cloud data, obtaining built-in structure data corresponding to each independent component in a component type with a built-in structure based on the design drawing of the civil engineering structure and the component modeling parameters, and generating a three-dimensional entity model of the civil engineering structure based on the component modeling parameters and the built-in structure data. The invention solves the defects of low efficiency, strong subjectivity and difficulty in reflecting the actual deformation of the structure of the traditional manual modeling, can accurately reflect the current situation of the target structure, and has important application value in the fields of finite element analysis and the like.

Inventors

HU HAO
XU YUEKAI
LI XIAOYA
YAN YUXUAN
ZHANG YONG
Kou jing
Shu Jiangpeng
Dong Tengfang
ZHOU LUJING
MA HAIBO

Assignees

浙江省交通运输科学研究院

Dates

Publication Date: 20260512
Application Date: 20260416

Claims (9)

1. The civil engineering structure modeling method based on the point cloud is characterized by comprising the following steps of: acquiring real scanning point cloud data and a design drawing of a civil engineering structure; performing multi-category semantic segmentation on the real scanning point cloud data by adopting a trained PointNet ++ network so as to divide point cloud data sets corresponding to various component types; re-dividing the point cloud data sets corresponding to each component type by adopting a density clustering algorithm to obtain independent point cloud data corresponding to each independent component; Acquiring component modeling parameters corresponding to each independent component based on the independent point cloud data corresponding to each independent component; Acquiring built-in structure data corresponding to each independent component in the component type with built-in structure based on a design drawing of the civil engineering structure and component modeling parameters corresponding to each independent component; And generating a three-dimensional solid model of the civil engineering structure based on the component modeling parameters corresponding to each independent component and the built-in structure data corresponding to each independent component in the component type with the built-in structure.
2. The method for modeling a civil engineering structure based on a point cloud as claimed in claim 1, wherein said PointNet ++ network adopts a multi-scale grouping strategy.
3. The civil engineering structure modeling method based on point cloud as claimed in claim 1, wherein the PointNet ++ network training method comprises the following steps: Acquiring a real scanning point cloud data set and a three-dimensional building information model of a training sample structure, and generating a plurality of virtual synthesized point cloud data sets based on the three-dimensional building information model and a plurality of different point cloud density ratios, wherein the point cloud density ratios represent the ratio of the point cloud densities of at least two different parts in the training sample structure; Respectively acquiring a training set and a verification set of each corresponding to-be-trained PointNet ++ network based on each virtual synthesized point cloud data set, respectively training each corresponding to-be-trained PointNet ++ network, and taking the point cloud density ratio of each corresponding virtual synthesized point cloud data set as a super parameter for each to-be-trained PointNet ++ network; Based on the real scanning point cloud data set, acquiring test sets of each PointNet ++ network to be trained, testing each PointNet ++ network after training, and taking the PointNet ++ network after training with the best test result as the PointNet ++ network after training.
4. The method for modeling a civil engineering structure based on a point cloud as claimed in claim 1, wherein said density clustering algorithm uses a DBSCAN algorithm.
5. The method for modeling a civil engineering structure based on a point cloud according to claim 1, wherein the step of obtaining the component modeling parameters corresponding to each independent component based on the independent point cloud data corresponding to each independent component includes: And for the cylindrical component type, respectively performing cylindrical fitting by adopting a RANSAC algorithm based on the independent point cloud data corresponding to each independent component of the component type, and obtaining the component modeling parameters corresponding to each independent component of the component type.
6. The method for modeling a civil engineering structure based on a point cloud as claimed in claim 5, wherein said obtaining the modeling parameters of each individual member based on the respective independent point cloud data of each individual member includes: For any independent component in the point cloud data set of the component type with the complex section, acquiring a section data point set of the independent component based on independent point cloud data of the independent component; Acquiring a boundary point set of the independent component by adopting an ordered concave-packed algorithm based on the section data point set of the independent component; acquiring a closed boundary point sequence by adopting a distance iterative algorithm based on the boundary point set of the independent component; Removing non-corner points in the boundary point sequence based on a preset angle threshold value to obtain component modeling parameters of the independent component; repeating the steps until the component modeling parameters corresponding to each independent component of the component type with the complex section are obtained.
7. The method for modeling a civil engineering structure based on a point cloud as claimed in claim 1, wherein the step of obtaining the built-in structure data corresponding to each independent component in the type of the component having the built-in structure based on the design drawing of the civil engineering structure and the component modeling parameters corresponding to each independent component, includes: establishing a drawing plane coordinate system based on a design drawing, and acquiring a coordinate point set of a corresponding built-in structure in the design drawing; For any independent component in the component types with built-in structures, a projection plane coordinate system is established based on component modeling parameters of the independent component; acquiring coordinate transformation parameters of the independent component from a drawing plane coordinate system to a projection plane coordinate system by adopting a Rodrigues algorithm; Transforming the coordinate point set of the independent component based on the corresponding coordinate transformation parameters to obtain a mapping point set of the independent component, and re-projecting the mapping point set to a projection plane coordinate system to correct coordinate transformation errors; obtaining built-in structure data of the independent component based on the corrected mapping point set of the independent component; repeating the steps until the built-in structure data corresponding to each independent component in the component type with the built-in structure is obtained.
8. The civil engineering structure modeling system based on the point cloud is characterized by comprising an acquisition unit, a segmentation unit, a construction parameter extraction unit, a built-in parameter reconstruction unit and a modeling unit: The acquisition unit acquires real scanning point cloud data and a design drawing of the civil engineering structure; The segmentation unit adopts a trained PointNet ++ network to carry out multi-category semantic segmentation on the real scanning point cloud data so as to divide point cloud data sets corresponding to various component types; The construction parameter extraction unit is used for respectively acquiring component modeling parameters corresponding to each independent component based on the independent point cloud data corresponding to each independent component; the built-in parameter reconstruction unit acquires built-in structure data corresponding to each independent component in the component type with built-in structure based on a design drawing of the civil engineering structure and component modeling parameters corresponding to each independent component; The modeling unit generates a three-dimensional solid model of the civil engineering structure based on the component modeling parameters corresponding to each independent component and built-in structure data corresponding to each independent component in the component type with built-in structure.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1-7 when the computer program is executed.

Description

Civil engineering structure modeling method and system based on point cloud and electronic equipment Technical Field Embodiments of the present specification relate to the field of three-dimensional modeling technology, and in particular, to optimization of a method for modeling a civil engineering structure based on a point cloud. Background In the field of civil engineering, a method for performing finite element analysis and evaluation on a structure based on a solid model is widely applied to research related mechanical and structural problems. However, the conventional modeling method relies on manual operations, and when a structural entity undergoes unknown deformation such as differential settlement and concrete shrinkage, the manual modeling sometimes causes deviation of the structure in finite element analysis. The point cloud is used as a three-dimensional data point set which can effectively reflect the current appearance characteristics (such as component deformation, dimension length, space position and the like) of the civil engineering structure, and can well acquire and reflect the current deformation data of the structure. The method has great application prospect in constructing a high-quality civil engineering solid model. However, because of the characteristics of irregular, disordered and the like of the point cloud data, the modeling data can only be extracted in an inefficient and subjective mode of manually picking up characteristic points in most application scenes at present. Therefore, a method for automatically, efficiently and accurately constructing a three-dimensional solid model from point cloud data is needed to solve the defects of low efficiency, strong subjectivity and difficulty in accurately reflecting the actual deformation of a structure in the prior art. Disclosure of Invention The embodiment of the specification provides a civil engineering structure modeling method, a system and electronic equipment based on point cloud, establishes a set of new strategy for automatic conversion from the point cloud to a three-dimensional entity model, and overcomes the defects that in the prior art, the modeling based on the point cloud is dependent on manpower, has low efficiency and strong subjectivity, and is difficult to accurately reflect the actual deformation of a structure. The technical scheme is as follows: In a first aspect, an embodiment of the present disclosure provides a method for modeling a civil engineering structure based on a point cloud, including the steps of: acquiring real scanning point cloud data and a design drawing of a civil engineering structure; performing multi-category semantic segmentation on the real scanning point cloud data by adopting a trained PointNet ++ network so as to divide point cloud data sets corresponding to various component types; re-dividing the point cloud data sets corresponding to each component type by adopting a density clustering algorithm to obtain independent point cloud data corresponding to each independent component; Acquiring component modeling parameters corresponding to each independent component based on the independent point cloud data corresponding to each independent component; Acquiring built-in structure data corresponding to each independent component in the component type with built-in structure based on a design drawing of the civil engineering structure and component modeling parameters corresponding to each independent component; And generating a three-dimensional solid model of the civil engineering structure based on the component modeling parameters corresponding to each independent component and the built-in structure data corresponding to each independent component in the component type with the built-in structure. Preferably, the PointNet ++ network adopts a multi-scale grouping strategy. As a preferred solution, the PointNet ++ training method of the network includes the following steps: Acquiring a real scanning point cloud data set and a three-dimensional building information model of a training sample structure, and generating a plurality of virtual synthesized point cloud data sets based on the three-dimensional building information model and a plurality of different point cloud density ratios, wherein the point cloud density ratios represent the ratio of the point cloud densities of at least two different parts in the training sample structure; Respectively acquiring a training set and a verification set of each corresponding to-be-trained PointNet ++ network based on each virtual synthesized point cloud data set, respectively training each corresponding to-be-trained PointNet ++ network, and taking the point cloud density ratio of each corresponding virtual synthesized point cloud data set as a super parameter for each to-be-trained PointNet ++ network; Based on the real scanning point cloud data set, acquiring test sets of each PointNet ++ network to be trained, testing each PointNet ++ netw