Abstract: The visualization of 3D city scenes is a key technology in digital cities, and it is also an important link in the provision of various smart services in smart cities. However, the amount and complexity of data in 3D city models grows geometrically, which imposes a heavy burden on data organization and visualization of urban 3D scenes, in the current city-level visualization platform based on game engine and geographic information system technology, this kind of data is difficult to load and visualize at one time. Aiming at this problem, this paper proposes a method for efficient visualization of 3D urban scenes that integrates hierarchical levels of detail(HLOD) and potentially visible sets(PVS), which realizes the efficiency of dynamic scheduling and loading of large-scale complex 3D urban scenes and the visual consistency of 3D visualization. A case study is carried out in the Qianhai area of Shenzhen. The results show that the average occlusion removal rate of the potentially visible sets calculation method based on adaptive partition proposed in this paper is as high as 76.6%, which can greatly remove and reduce the amount of model data loaded into a 3D city scene; Integrating the dynamic scheduling of HLOD and PVS, the rendering visualization frame rate is basically maintained at around 60 FPS, which can meet the efficient and smooth visualization experience of 3D urban scenes.

Key words: smart city, 3D visualization, adaptive partitioning, potential visible set, HLOD