Research On Single Building Extraction For Dense Matching Point Cloud

As a new 3D spatial data,dense matching point cloud can provide data support for the largescale and low-cost acquisition of single buildings.However,due to the uneven image quality and the limitation of matching algorithm,there are a lot of noise points and artifacts in the generated point cloud data.In addition,there are vegetation point clouds closely connected with buildings in the point cloud.All these factors will affect the effect of building monomer extraction,and even cause statistical errors of urban housing information.To solve the above problems,a single building extraction method for dense matching point cloud is proposed.Firstly,the spatial distribution characteristics are calculated based on the point cloud-scale information and combined with the RGB information selected by the voting mechanism,the nonbuilding point clouds dominated by vegetation points are removed to reduce the influence of irrelevant point clouds.Secondly,the octree principle is used to voxel the point cloud to complete the down sampling,to solve the problem of extraction efficiency caused by large data volume and uneven density noise points.Thirdly,in order to solve the problem of inaccurate calculation of the traditional rough point cloud surface normal vector,the moving least square vector estimation method is used to extract the robust normal vector.Thirdly,in order to solve the problem of inaccurate calculation of the traditional rough point cloud surface normal vector,the moving least square vector estimation method is used to extract the robust normal vector.In the experiment,the dense matching point cloud data of Wuhan Xinzhou district and Switzerland were used to analyze the results.The filtering accuracy of the non-building point cloud reached 92.26% and 96.04%respectively,which improved the effect by 5.03% and 4.51% compared with the traditional single exponential filtering method.The average accuracy of building monomer segmentation reaches97.85%.Compared with traditional region growth algorithm,building monomer algorithm based on histogram,DBSCAN clustering algorithm,and ST-DBSCAN clustering algorithm,the proposed algorithm has higher accuracy and better effect in extracting single buildings.This thesis has 27 pictures,9 tables,and 51 references.