Comprehensive Evaluation of Modeling and Surface Simplification Methods for 3D Building Reconstruction from Dense Point Cloud

3D building reconstruction is a mathematic model and representation of 3D surfaces for building details in urban areas. There are many methods for 3D modeling such as Image Based Rendering (IBR), Image Based Modeling (IBM) and Range Based Modeling (RBM). These methods use generated 3D point cloud from different techniques sources such aerial laser scanners and photogrammetry multi view imageries. In this paper, 3D model generation methods based on triangulation algorithms such as Poisson, ball-pivoting and volumetric triangulation using Marching Cubes (MC) are evaluated using a raw dense point cloud. Also two mesh simplification methods called clustering decimation and quadric edge collapse are used to improve the quality of triangulated models with decrease the surface and vertex numbers. A geometric metric called Hausdorff distance is used for comparison of each model with a reference. The results show that the accuracy of generated 3D model based on volumetric triangulation method using Marching Cubes (MC) is better than other methods. Also, quadric edge collapse method can simplified 3D models better than clustering decimation method.