Single Extraction And 3D Modeling Of Urban Objects Based On Airborne LiDAR Point Cloud

The single extraction and 3D modeling of urban objects is not only conducive to urban planning and management,but also makes urban services smarter and more convenient.The modeling of the building unit in the urban area is divided into two stages,one is to extract the building unit in the urban area,and the other is to perform three-dimensional modeling of the extracted building unit.Due to the huge amount of point cloud data,the various types of features,and the significant differences in the scale,shape,and location distribution of the building units,it is very challenging to accurately extract the independent range of the building units.One of the difficulties faced by single-unit 3D modeling is that facing building units with different roof structures and different shapes,there is a lack of a 3D modeling method that does not rely on model libraries or human prior knowledge.Another challenge for building single building modeling is the top-down scanning method of airborne point clouds,which leads to sparse building point cloud facades,cracks or even missing,which affects the 3D modeling effect.In view of the above key issues,the main research contents of this article are as follows:In view of the large amount of point cloud data in the airborne LiDAR urban area and the chaotic and diverse types of ground features,the 3D point cloud data is not directly processed,but the 2D raster data after the point cloud dimension reduction is used to locate and locate buildings.The extraction strategy not only reduces the complexity of the calculation,but also eliminates the lengthy process of 3D point cloud labeling;for the location distribution,scale and shape of the building is difficult to predict,and the adhesion of other features such as tall vegetation to adjacent buildings.Detecto RS,an instance segmentation algorithm with high sensitivity to target objects of different scales,is used to obtain the initial results of the extraction of individual buildings.Aiming at the mask overlap phenomenon in the initial results,an edge optimization strategy based on mask fusion is proposed.The refined extraction result of the building monomer is obtained.When 3D modeling is performed on the point cloud of a single building,the complexity of the roof structure between buildings is different,and the shape difference is obvious.The existing 3D modeling method relies on problems such as model library or a large amount of data driving.It can be used directly from the point Point2 Mesh,an unsupervised learning network that learns shape information from cloud data,carries out a three-dimensional modeling method for individual buildings;in view of the impact of local sparseness and defects of buildings on three-dimensional modeling,a point based on the repair of bottom holes is proposed.The cloud completion strategy obtains better closed point cloud data,thereby improving the 3D modeling effect of building point clouds and realizing 3D surface modeling of various complex building point clouds.