Research On Construction Accuracy Detection Of Prefabricated Structure Based On Realistic 3D Model

In modern transportation engineering,more and more bridges and transportation buildings are adopting prefabricated structures to achieve efficient,green,and environmentally friendly purposes.The precision of component fabrication and construction and installation errors during the assembly structure construction process directly affect the quality of the overall structure.With the comprehensive application of BIM technology,the use of reverse-constructed reality 3D models to compare and detect component dimensional accuracy and construction installation deviation has gradually become a research hotspot.This thesis,based on the dense point cloud established by fusing aerial and ground-source heterogeneous images,uses a surface mesh stitching algorithm to establish a high-precision3 D model,which efficiently and accurately detects assembly structure components and construction deviations compared with the forward design model.The main research contents of this thesis are as follows:1.Aerial and ground-source heterogeneous image fusion technology generates dense point clouds.Unmanned aerial vehicles and ground cameras collect aerial and ground image data,and by using transition-linked images,the heterogeneous image data are fused to effectively establish a more complete and accurate dense point cloud.This solves the problem of mutual obstruction of components on the assembly structure construction site,making it difficult to establish a clear and complete dense point cloud,thereby improving the accuracy and stability of the point cloud.2.A surface mesh stitching algorithm is used to establish a high-precision 3D model.The dense point cloud is subjected to statistical filtering and denoising and voxel grid downsampling.The point cloud smooth and sharp areas are partitioned using surface variation,and the Poisson algorithm is used to optimize the reconstruction of the smooth area point cloud,while the greedy projection triangulation algorithm is used to optimize the reconstruction of the sharp area point cloud.Finally,this thesis proposes surface mesh stitching algorithm is used to stitch the two optimized surfaces to establish a high-precision3 D model.The results show that the maximum positioning error and average median error of the 3D model established by the surface mesh stitching algorithm are 2.029 mm and1.799 mm,respectively,which is superior to the accuracy of the Poisson algorithm(9.508 mm,8.243mm)and the greedy projection triangulation algorithm(5.525 mm,3.921mm).3.Forward and reverse 3D model spatial registration and construction precision detection.Using the forward design 3D model as a benchmark,the reality 3D model mentioned above is registered for the second time in space.The dimensional accuracy and construction precision of the assembly structure components are obtained through the distance and feature line extraction method of registration points.The experiment showed that the maximum deviation of component dimensions was 1.921 mm,and the maximum deviations in the detection of axis position,verticality,and adjacent component flatness were2.809 mm,2.731 mm,and 2.351 mm,respectively,which were better than the results obtained by traditional detection methods.Through the actual application of a prefabricated structure project,the maximum deviation of the prefabricated stair component size detection was 2.698 mm,and the maximum deviations in the detection of axis position,verticality,and adjacent component flatness of 11 prefabricated wall panels were 1.769 mm,2.117 mm,and 2.201 mm,respectively,all of which comply with the requirements of the "Acceptance Specification for Construction Quality of Concrete Structures" with a permissible deviation of 5mm.