Abstract

Point cloud shape completion aims to reconstruct complete point clouds from partial point clouds. The denser point clouds are, the richer information they contain. Different from existing methods that pay more attention to sparse completion of point clouds and global feature information of partial point clouds, this paper proposes a novel dense point cloud completion network called N-DPC, which combines self attention unit with the fusion of local feature and global feature information. First, we apply self attention unit to point clouds' global feature extraction to make it lay emphasis on the dependency between different points. Second, we adopt the method of multi-stage completion where we obtain coarse point clouds at first stage and combine local and global information to achieve the completion of dense point clouds. Quantitative and qualitative evaluations of experiments demonstrate that the proposed method has achieved better performance on ShapeNet dataset compared with existing state-of-the-art point cloud completion works and shows a good robustness for different missing ratios of point clouds. Additionally, the proposed N-DPC is valid for real point clouds on KITTI dataset.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call