Fast growing demand for large scale 3D reconstruction is accelerated by the popularization of photography drones and cameras in mobile phones for the past few years. Capturing of high quality photos has never been made easier and this pushes the reconstruction scale to a new level, introducing new challenges to the mesh processing techniques.

This thesis proposed a level-of-detail(LOD) mesh generation pipeline that is robust and scalable to the ever growing data size. It improved three major mesh operations which are used heavily in the pipeline, fixing challenging mesh artifacts, merging meshes without overlapping requirement or generating degenerated triangles and adapting a generic mesh simplification algorithm for urban scene. Experiment on a large scale data set demonstrate...[ Read more ]

Fast growing demand for large scale 3D reconstruction is accelerated by the popularization of photography drones and cameras in mobile phones for the past few years. Capturing of high quality photos has never been made easier and this pushes the reconstruction scale to a new level, introducing new challenges to the mesh processing techniques.

This thesis proposed a level-of-detail(LOD) mesh generation pipeline that is robust and scalable to the ever growing data size. It improved three major mesh operations which are used heavily in the pipeline, fixing challenging mesh artifacts, merging meshes without overlapping requirement or generating degenerated triangles and adapting a generic mesh simplification algorithm for urban scene. Experiment on a large scale data set demonstrated its robustness and scalability.