Towards real-time incremental surface reconstruction from streaming point cloud

Abstract

Surface reconstruction, also known as mesh reconstruction, is a critical step in 3D reconstruction pipeline that generates 3D models from a series of images of a specific objects. The input of the surface reconstruction is a sparse 3D point cloud obtained from Structure-from-Motion (SfM) and the output is a 3D model represented by a mesh. In general, this process is implemented off-line because it requires a cost calculation of optimizing labelling energy and cannot insert or remove points dynamically. In this thesis, a new reconstruction method is implemented which can incrementally extract surface from tetrahedra after the triangulation from streaming point cloud. A novel energy function is harnessed in order to reduce the time complexity and without losing the accuracy comparing to state-of-the-art method. By utilizing this energy function and dynamic version of graph cut algorithm, the reconstruction process can be achieved in real-time and dynamic manner. With this implementation, 3D reconstruction can be applied in the real-time applications such as Simultaneous Localization and Mapping (SLAM).

Date of Award	2020
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology