Research On Buildings Structure Perception And Vectorization Based On Dense Matching Point Cloud

As an important part of urban geographic information,building vector information is widely used in urban planning,municipal management,public safety and other fields.In the fields of fine analysis of buildings and urban renewal,higher requirements are put forward for the vectorization of buildings.The vectorization results of building structures which contain the information of partial structure identification such as balconies and piaolu play an important role.As a new type of spatial data with low cost and high precision,dense matching point cloud plays an increasingly important role in the construction and development of digital city.In order to solve the low efficiency problem of vectorization of building structure through traditional manual drawing,this paper takes dense matching point cloud as data source to study the key technologies in the reconstruction of building structural vector model.The main research contents are as follows:(1)For the extraction of monomer point cloud of building structure in dense building area,On the basic of ground filtering and clustering after horizontal point cloud extraction,the method projects all the point cloud clusters into the grid and removes non-roof segments based on building fa?ade and clusters' geometrical characteristic.The topological relationship between roof surfaces is further calculated based on the raster image and structured segmentation was carried out to realize the extraction of monomer point cloud of building structure.(2)As for the vector contour reconstruction for buildings,the convex hull algorithm based on neighborhood constraints and the linear growth algorithm based on least square are used to extract contours from monomer point cloud of building structure,and W-kmeans algorithm is used to correct the direction of contours.The algorithm is optimized by adding angle constraint to satisfy the vector contour reconstruction of irregular buildings.(3)In terms of local structure identification,The eaves are corrected based on the contours of the building profile to avoid the influence of the eaves structure on the vectorization results.In view of the identification of the building typical local structure,such as balcony and flutter.The profile feature tracking method was used to extract the profile structure of the building structure,and the automatic identification of the balcony and flutter was realized by combining the topological contrast analysis method.The research shows that,through the point cloud extraction method of monomer point cloud of building structure in this paper.In the dense building scene,most of the buildings can be distinguished and building structure can be segmented correctly,which provides a good data basis for the vectorization of the building structure in the later stage.The vector contour reconstruction method and the local structure identification method in this paper can reflect the real vector structure characteristics of buildings,and can extract the vector information of buildings quickly,accurately and automatically.