Distinguishing and Matching-Aware Unsupervised Point Cloud Completion

doi:10.1109/TCSVT.2023.3250970

CORC > 力学研究所 > 中国科学院力学研究所

	Distinguishing and Matching-Aware Unsupervised Point Cloud Completion
	Xiao, Haihong 5; Li YQ(李玉琼)4; Kang, Wenxiong 2,3,5; Wu, Qiuxia 1
刊名	IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
	2023-09-01
卷号	33 期号:9 页码:5160-5173
关键词	Deep learning point cloud completion 3D vision
ISSN号	1051-8215
DOI	10.1109/TCSVT.2023.3250970
通讯作者	Li, Yuqiong(liyuqiong@imech.ac.cn) ; Kang, Wenxiong(auwxkang@scut.edu.cn)
英文摘要	Real-scanned point clouds are often incomplete due to occlusion, light reflection and limitations of sensor resolution, which impedes the related progress of downstream tasks, e.g., shape classification and object detection. Although there has been impressive research progress on the point cloud completion topic, they rely on the premise of extensive paired training data. However, collecting complete point clouds in some specified scenarios is labor-intensive and even impractical. To mitigate this problem, we propose DMNet, a distinguishing and matching-aware unsupervised point cloud completion network. Our work belongs to the group of unsupervised completion methods but goes beyond previous studies. Firstly, we propose a distinguishing-aware feature extractor to learn discriminable semantic information for different instances, simultaneously enhancing the robust invariant representation under noise disturbances. Secondly, we design a hierarchy-aware hyperbolic decoder to recover the complete geometry of point clouds, which not only can capture the implicit hierarchical relationships in data but also has an explicit extended nature. Finally, we develop a matching-aware refiner to eliminate noise points via aligning the topology structure of the input and predicted partial point clouds. Extensive experiments on MVP, Completion3D and KITTI datasets prove the effectiveness of our method, which performs favorably over state-of-the-art methods both quantitatively and qualitatively.
分类号	一类
资助项目	Youth Innovation Promotion Association of the Chinese Academy of Sciences[2018024] ; National Natural Science Foundation of China[61976095] ; National Natural Science Foundation of China[61575209] ; Experiments for Space Exploration Program, Qian Xuesen Laboratory, China Academy of Space Technology[TKTSPY-2020-05-01]
WOS研究方向	Engineering
语种	英语
WOS记录号	WOS:001063316800053
资助机构	Youth Innovation Promotion Association of the Chinese Academy of Sciences ; National Natural Science Foundation of China ; Experiments for Space Exploration Program, Qian Xuesen Laboratory, China Academy of Space Technology
其他责任者	Li, Yuqiong ; Kang, Wenxiong
内容类型	期刊论文
源URL	[http://dspace.imech.ac.cn/handle/311007/92987]
专题	中国科学院力学研究所
作者单位	1.South China Univ Technol, Sch Software Engn, Guangzhou 510006, Peoples R China 2.Young Scholar Project Ctr, Pazhou Lab, Guangzhou 510335, Peoples R China; 3.South China Univ Technol, Sch Future Technol, Guangzhou 510641, Peoples R China; 4.Chinese Acad Sci, Key Lab Mech Fluid Solid Coupling Syst, Inst Mech, Beijing 100190, Peoples R China; 5.South China Univ Technol, Sch Automat Sci & Engn, Guangzhou 511442, Peoples R China;
推荐引用方式 GB/T 7714	Xiao, Haihong,Li YQ,Kang, Wenxiong,et al. Distinguishing and Matching-Aware Unsupervised Point Cloud Completion[J]. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY,2023,33(9):5160-5173.
APA	Xiao, Haihong,李玉琼,Kang, Wenxiong,&Wu, Qiuxia.(2023).Distinguishing and Matching-Aware Unsupervised Point Cloud Completion.IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY,33(9),5160-5173.
MLA	Xiao, Haihong,et al."Distinguishing and Matching-Aware Unsupervised Point Cloud Completion".IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY 33.9(2023):5160-5173.