In this paper, we propose a novel method to jointly solve scene layoutestimation and global registration problems for accurate indoor 3Dreconstruction. Given a sequence of range data, we first build a set of scenefragments using KinectFusion and register them through pose graph optimization.Afterwards, we alternate between layout estimation and layout-based globalregistration processes in iterative fashion to complement each other. Weextract the scene layout through hierarchical agglomerative clustering andenergy-based multi-model fitting in consideration of noisy measurements. Havingthe estimated scene layout in one hand, we register all the range data throughthe global iterative closest point algorithm where the positions of 3D pointsthat belong to the layout such as walls and a ceiling are constrained to beclose to the layout. We experimentally verify the proposed method with thepublicly available synthetic and real-world datasets in both quantitative andqualitative ways.
展开▼
