Partition-Based Point Cloud Completion Network with Density Refinement
In this paper, we propose a novel method for point cloud complementation called PADPNet. Our approach uses a combination of global and local information to infer missing elements in the point cloud. We achieve this by dividing the input point cloud into uniform local regions, called perceptual field...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-07-01
|
Series: | Entropy |
Subjects: | |
Online Access: | https://www.mdpi.com/1099-4300/25/7/1018 |
_version_ | 1797589403164475392 |
---|---|
author | Jianxin Li Guannan Si Xinyu Liang Zhaoliang An Pengxin Tian Fengyu Zhou |
author_facet | Jianxin Li Guannan Si Xinyu Liang Zhaoliang An Pengxin Tian Fengyu Zhou |
author_sort | Jianxin Li |
collection | DOAJ |
description | In this paper, we propose a novel method for point cloud complementation called PADPNet. Our approach uses a combination of global and local information to infer missing elements in the point cloud. We achieve this by dividing the input point cloud into uniform local regions, called perceptual fields, which are abstractly understood as special convolution kernels. The set of point clouds in each local region is represented as a feature vector and transformed into N uniform perceptual fields as the input to our transformer model. We also designed a geometric density-aware block to better exploit the inductive bias of the point cloud’s 3D geometric structure. Our method preserves sharp edges and detailed structures that are often lost in voxel-based or point-based approaches. Experimental results demonstrate that our approach outperforms other methods in reducing the ambiguity of output results. Our proposed method has important applications in 3D computer vision and can efficiently recover complete 3D object shapes from missing point clouds. |
first_indexed | 2024-03-11T01:07:12Z |
format | Article |
id | doaj.art-e7216e64514849ee86feebfa26dca2f8 |
institution | Directory Open Access Journal |
issn | 1099-4300 |
language | English |
last_indexed | 2024-03-11T01:07:12Z |
publishDate | 2023-07-01 |
publisher | MDPI AG |
record_format | Article |
series | Entropy |
spelling | doaj.art-e7216e64514849ee86feebfa26dca2f82023-11-18T19:13:32ZengMDPI AGEntropy1099-43002023-07-01257101810.3390/e25071018Partition-Based Point Cloud Completion Network with Density RefinementJianxin Li0Guannan Si1Xinyu Liang2Zhaoliang An3Pengxin Tian4Fengyu Zhou5School of Electrical Engineering, Academy of Information Sciences, Shandong Jiaotong University, Jinan 250357, ChinaSchool of Electrical Engineering, Academy of Information Sciences, Shandong Jiaotong University, Jinan 250357, ChinaSchool of Electrical Engineering, Academy of Information Sciences, Shandong Jiaotong University, Jinan 250357, ChinaSchool of Electrical Engineering, Academy of Information Sciences, Shandong Jiaotong University, Jinan 250357, ChinaSchool of Electrical Engineering, Academy of Information Sciences, Shandong Jiaotong University, Jinan 250357, ChinaSchool of Control Science and Engineering, Shandong University, Jinan 250012, ChinaIn this paper, we propose a novel method for point cloud complementation called PADPNet. Our approach uses a combination of global and local information to infer missing elements in the point cloud. We achieve this by dividing the input point cloud into uniform local regions, called perceptual fields, which are abstractly understood as special convolution kernels. The set of point clouds in each local region is represented as a feature vector and transformed into N uniform perceptual fields as the input to our transformer model. We also designed a geometric density-aware block to better exploit the inductive bias of the point cloud’s 3D geometric structure. Our method preserves sharp edges and detailed structures that are often lost in voxel-based or point-based approaches. Experimental results demonstrate that our approach outperforms other methods in reducing the ambiguity of output results. Our proposed method has important applications in 3D computer vision and can efficiently recover complete 3D object shapes from missing point clouds.https://www.mdpi.com/1099-4300/25/7/1018convolutional neural networkspoint cloud completiongriddingradargeometric density |
spellingShingle | Jianxin Li Guannan Si Xinyu Liang Zhaoliang An Pengxin Tian Fengyu Zhou Partition-Based Point Cloud Completion Network with Density Refinement Entropy convolutional neural networks point cloud completion gridding radar geometric density |
title | Partition-Based Point Cloud Completion Network with Density Refinement |
title_full | Partition-Based Point Cloud Completion Network with Density Refinement |
title_fullStr | Partition-Based Point Cloud Completion Network with Density Refinement |
title_full_unstemmed | Partition-Based Point Cloud Completion Network with Density Refinement |
title_short | Partition-Based Point Cloud Completion Network with Density Refinement |
title_sort | partition based point cloud completion network with density refinement |
topic | convolutional neural networks point cloud completion gridding radar geometric density |
url | https://www.mdpi.com/1099-4300/25/7/1018 |
work_keys_str_mv | AT jianxinli partitionbasedpointcloudcompletionnetworkwithdensityrefinement AT guannansi partitionbasedpointcloudcompletionnetworkwithdensityrefinement AT xinyuliang partitionbasedpointcloudcompletionnetworkwithdensityrefinement AT zhaoliangan partitionbasedpointcloudcompletionnetworkwithdensityrefinement AT pengxintian partitionbasedpointcloudcompletionnetworkwithdensityrefinement AT fengyuzhou partitionbasedpointcloudcompletionnetworkwithdensityrefinement |