Research On Structure Scanning Point Cloud Feature Analysis And Cavity Repair Technology Considering Multi-source 3D Data Acquisitio

As one of the best methods for obtaining the geometric shape information of complex structures,3D point cloud data acquisition can be achieved through laser scanning or photogrammetric methods.However,due to differences in sensor mechanisms and field acquisition schemes,point cloud acquisition methods based on different platforms can utilize their platform advantages to obtain different point clouds for the target under different measurement scenarios and environmental conditions.However,this also results in data differences in quantity,quality,types of information contained,and geometric features between heterogeneous point clouds.Single-source point clouds cannot meet the application requirements of different scenarios,and there is currently no data acquisition method that can simultaneously take into account the advantages of various types of point clouds.Moreover,although data fusion can be performed between heterogeneous point clouds,if various point cloud data are directly fused,the resulting point cloud will not only contain a large amount of redundant data but may also suffer from quality degradation.Additionally,it may not perform well in subsequent processing applications of the point cloud.Therefore,in order to achieve the complementary advantages of multiple sources of point cloud data and to provide high-quality basic 3D point cloud data for structure measurement and modeling,this study conducted a comparative analysis of the same heterogeneous structure point cloud data obtained by handheld three-dimensional laser scanning technology(HLS),terrestrial three-dimensional laser scanning technology(TLS),and unmanned aerial vehicle systems(UAS)low-altitude photogrammetric technology from three perspectives including the different sensor point cloud acquisition mechanisms,original data quality,and multi-scale geometric features.Based on the analysis of the differences,TLS point cloud was selected as the base point cloud,and the point clouds obtained by the other two methods were used as patch data to repair the holes in the TLS point cloud.The specific research content is as follows:(1)This study compared the quality of the original point cloud data from different heights in terms of the amount of point cloud data obtained and the spatial information precision,respectively,calculated point cloud roughness,curvature,FPFH local features,and VFH global features.Quantitative evaluation of the multi-scale geometric feature differences of heterogeneous point clouds was performed using indicators such as M3C2 distance,discrete Fréchet distance,and standard Euclidean distance.(2)Based on the analysis of point cloud feature differences,the TLS point cloud,which had better data quality but still had some point cloud holes,was selected as the base point cloud.Using the M3C2 distance algorithm,significant point changes between the base point cloud and the HLS point cloud were extracted,and the difference in single-point geometric features between the two types of point clouds was obtained.Based on the core density distribution of feature differences,feature points were selected from the HLS point cloud and fused with the TLS point cloud,thereby enriching the feature point count of the TLS point cloud.For repairing the top hole,this study used methods such as grid projection and Euclidean cluster segmentation to extract repair point clouds from the UAS imaging point cloud and repair the top hole in the TLS point cloud.At the same time,this study used a small amount of point cloud remaining in the TLS point cloud side hole as seed points to perform nearest neighbor search on the HLS point cloud data to repair the side hole of the TLS point cloud.Through research,it was found that there were significant differences in the quality of original point cloud data and multi-scale geometric features between the same heterogeneous point clouds.Both laser point clouds have high spatial information accuracy and data acquisition efficiency,but they are susceptible to data missing.Additionally,different laser point clouds obtained by different data acquisition mechanisms have significantly different geometric feature information.UAS imaging point clouds can effectively supplement the data missing in laser point clouds,and the two types of point clouds have good complementarity.Furthermore,compared to directly fusing multiple source point clouds,the TLS point cloud repaired by the holefilling method in this article can obtain complete structural 3D point cloud information containing heterogeneous point clouds without increasing data redundancy.The results of this study provide high-quality basic point cloud data for subsequent data processing applications and provide a new solution for the comprehensive utilization of target multi-source 3D point cloud data.