Organization And Scheduling Method Of Complex 3D Urban Scene Data Driven By Visual Perception

At present,the rapid development of data collection and modeling technology provide a variety of data products for smart city,including large-scale digital surface model data,point cloud data,3D building model data,etc.The exponential growth of the three-dimensional urban model data has brought a heavy burden to the data organization and management.In online real-time applications,this kind of data is difficult to transmit from the server side to the client at a time.The traditional division method based on area is difficult to be applied to large scale complex urban area,a self-adaptive organization and scheduling method for 3D urban scene driven by visual perception is proposed in this paper.The main contents of this research are organized as follows:(1)Firstly the background and significance of this paper are introduced,and then the situation and the development trend of the visualization of large-scale complex threedimensional urban scene data are summarized,finally the purpose of this study and research content is put forward;(2)Adaptive organization of complex three-dimensional urban scene data.Based on the characteristics of spatial data types,according to the distribution density of spatial data and coverage areas,three-dimensional city scene data is divided by different spatial levels and granularities using the spatial data partitioning method.Secondly,it traverses the quadtree from the bottom up,generates LODs for the intermediate nodes,and calculates the geometric errors at each level.The result of the adaptive quadtree partitioning is produced as a threedimensional tile,which generates a continuous hierarchical LOD for the urban scene;(3)3D tile model that combines adaptive partitioning and multiple levels of detail is designed.The design of 3D tile model is divided into two parts: index file and tile file.The index file organizes the index of all city tile files(b3dm),and each tile stores part of the city's data;(4)Viewpoint-related HLOD scheduling is based on the current viewpoint information.Within a given screen error threshold range,the scene is traversed gradually,and the HLOD that satisfies the error requirement is dynamically selected.The selection of the HLOD depends not only on the distance of the line of sight,but also includes the visual perception parameters of the 3D geometric features(shape,size,height)of the building model.Based on these impact factors,the visual perception evaluation model constructed in this paper can quickly select the correct tile.In order to verify the effectiveness of the proposed method for adaptive organization and dynamic scheduling of complex 3D urban scene data,a prototype system was developed in this paper.Large-scale visualization experiments of 3D models and large-scale visualization experiments of digital surface models were performed.Experiments show that based on the dynamic scheduling of viewpoint-related HLOD,the rendering frame rate is always maintained at about 40 fps,which can guarantee the high efficiency and visual consistency of the dynamic scheduling of complex 3D urban scene data in the network environment.