Data-driven Reconstruction for Massive Buildings within Urban Scenarios: A Case Study

The reconstruction and visualization of urban environments is widely used in immersive simulation, but the massive data of buildings bring huge burden to the modelling and rendering process. This paper proposes an improved fast approach to build and visualize the scene graph based on the level of details (LOD). Firstly, an automatic reconstruction scheme of urban buildings is designed by using the public vector data set. Secondly, a tree-type organization and management structure is constructed for urban scenarios inspired by the idea of recursive division. Thirdly, a dynamic data scheduling strategy is developed with concerns of self-organizing the generated 3D massive model data. The developed approach is called Detail Segmentation of Scene Tree Method, namely DSSTM. Finally, experiments are conducted for performance comparison between the proposed DSSTM and a conventional method, while both are realized by using the open source library, OpenSceneGraph, http://www.openscenegraph.org/. The experimental results of reconstructing Changsha City have verified that the proposed method is effective and efficient for practical applications of visualization on large-scale urban with massive buildings. The rendering speed is increased to 4~7 times of the conventional method, the dynamic loading time is shortened by 90%.