Dimensional Quality Assessment And BIM Modeling Of Precast Concrete Box Beam Based On 3D Point Clouds

In order to promote the development of prefabricated construction technology and component quality assessment method in bridge engineering,enough combination of construction industry,information and industrialization is significantly necessary.The performance of prefabricated buildings depends heavily on the quality of the prefabricated components.Traditional dimensional measurement methods rely on tools such as laser rangefinder,water level,tape,which is time-consuming and less accurate.The measurement results are interfered by human’s subjective factors and the overall 3D geometry features cannot be evaluated.3D laser scanning technology is non-contact and highly accurate,turning traditional single point measurement mode into surface measurement mode.Dimension extraction of components can be realized with effective point cloud data processing algorithm,which is of great significance to promote the technology progress of prefabricated buildings.This study reports an intelligent dimension measurement and as-built BIM modeling technology of precast box beam based on 3D point clouds.Firstly,the pre-processing of raw point cloud including denoising and downsizing is performed.Secondly,the surface segmentation,edges extraction and corner points computation are performed.Next,the incomplete surface of the point cloud is repaired and reconstructed.Then the point clouds in local sections of the component is obtained by slice data projection.Then the dimensions are computed based on the spatial relations of geometric features of surface,line and corner points.A comparison of dimensions extracted from point clouds and manual measurement was conducted on eight precast prestressed concrete box beam,and the results show that 80.3% of the 132 pairs of comparison data have an absolute difference less than 10 mm,which validates the proposed method in this study.With comparison of dimensions extracted from point clouds and design values,the beam dimensional quality can be evaluated.At the same time,the bending and torsion deformation state can be assessed.Finally,Revit secondary development technology is used in automatic as-built BIM modeling is realized based on corner points of cross sections from point clouds.The obvious density inhomogeneity of the actual scanning point clouds of components in the construction site bring technical difficulties for data downsizing and surface segmentation.In order to preserve the points with relatively higher curvature characteristics and protect the points in the low-density area,this study propose a hybrid downsizing method.The core of the hybrid method is as follows.Firstly,the point clouds is downsized using 3D uniform voxel grid.Secondly,the 3D points are turned into 2D by slice projection method and the points with larger curvature are extracted.Then,the sampling rate is determined based on the density of point cloud to avoid holes caused by removing too many points in the low-density area.As for featuring surface segmentation,this study proposes a new criterion for the initial seed point selection of region growth method,which relies mainly on analyzing the uniformity of points.The experiment to determine optimal combination of threshold value and parameter is performed with points associated to the bottom of flange of the box beam.The proposed provides a new idea for the segmentation of non-uniform point cloud area.In this paper,Revit secondary development technology is used to transform the design BIM model of component into design point cloud based on Dynamo,which provides reference for component class identification and scanning point cloud registration.The as-built BIM model is generated based on the corner points of cross sections from point clouds.The model can represent the actual deformation of component,including bending and torsion.This study develops a dimension query external application,which take as-built BIM model as geometric information carrier,improving the interaction between model and information.