| The quality inspection and virtual pre-assembly of independent steel girder prefabricated components is an important part of modern bridge construction.The traditional manual method of component quality inspection is inefficient and timeconsuming,and does not meet the requirements of lean engineering construction.With the continuous improvement of 3D laser scanning technology and the popularization of applications in various fields,the use of 3D laser scanning technology to obtain refined 3D models of objects has become another revolution in surveying and mapping technology.In order to quickly and accurately inspect the quality of independent steel girder components and realize the digital assembly test of component models,this text uses the high-precision point cloud model of the bridge Ibeam as the experimental data,and proposes a semantic recognition algorithm for bridge component feature surfaces based on the point cloud model,automatically extract the defined component semantic features,and realize component part quality parameter detection algorithm and virtual pre-assembly according to the extracted feature semantic plane.The main work and achievements of this paper are as follows:(1)According to the bridge inspection engineering technical specifications and algorithm inspection standards,the steel plate girder data scanning scheme is designed,and the comprehensive surface information of the components is obtained by specifying the steel plate I-beam placement form and the location of the scanning station.In order to achieve accurate splicing of data between different sites and identification and calculation of special feature surfaces,target balls are placed at key parts of components as key feature points for algorithm identification,and a refined point cloud model of components with detection features is finally established.(2)Aiming at the problems of wide distribution of small parts and many surface fine structures in the component model,a complex scene segmentation algorithm based on super-voxel region clustering is adopted,and the component model is segmented by super-voxel and then region-grown and clustered to make the components with the same characteristics The obvious large-area combination is refused,the small and irrelevant feature bodies are removed,and the different structural areas of the components are effectively divided,and the RANSAC algorithm is used for comparative experiments to test the accuracy and efficiency of different algorithms.(3)According to the component quality inspection specification and the distribution of model components,define the feature semantic plane,combine the spatial geometric features of the model(density,normal vector,spatial connectivity)to establish semantic extraction rules,and extract the key feature semantic plane involved in steel beam component detection;secondly,combined with the slot expansion mechanism of Cloud Compare to realize the visual interaction and automatic processing of the model;finally,according to the steel I-beam quality inspection specification and assembly process.(5)Engineering application cases.An example is given to illustrate the application of the feature semantic surface in the dimension detection of steel beam members and the direction of virtual pre-assembly.The experimental results show that the algorithm can accurately extract the semantic plane of key features.In the application of parameter detection,the maximum absolute error of the beam height is1.65 mm,and the maximum absolute error of the camber is 5.8 mm,all of which meet the requirements of the specification.The algorithm can accurately and quickly identify the key features of components.The semantic feature surface provides a new idea for the quality and safety inspection of steel beam components. |
PDF Full Text Request