计算机科学 ›› 2022, Vol. 49 ›› Issue (8): 273-278.doi: 10.11896/jsjkx.210900023

• 人工智能 • 上一篇    下一篇

基于可变形图卷积的点云表征学习

李宗民1, 张玉鹏1, 刘玉杰1, 李华2   

  1. 1 中国石油大学(华东)计算机科学与技术学院 山东 青岛 266580
    2 中国科学院计算技术研究所 北京 100190
  • 收稿日期:2021-09-03 修回日期:2022-03-24 发布日期:2022-08-02
  • 通讯作者: 张玉鹏(S19070005@s.upc.edu.cn)
  • 作者简介:(lizongmin@upc.edu.cn)
  • 基金资助:
    国家重点研发计划(2019YFF0301800);国家自然科学基金(61379106);山东省自然科学基金(ZR2013FM036,ZR2015FM011)

Deformable Graph Convolutional Networks Based Point Cloud Representation Learning

LI Zong-min1, ZHANG Yu-peng1, LIU Yu-jie1, LI Hua 2   

  1. 1 College of Computer Science and Technology,China University of Petroleum,Qingdao,Shandong 266580,China
    2 Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100190,China
  • Received:2021-09-03 Revised:2022-03-24 Published:2022-08-02
  • About author:LI Zong-min,born in 1965,Ph.D,professor,is a member of China Computer Federation.His main research interests include computer graphics,picture processing and scientific computing visuali-zation.
    ZHANG Yu-peng,born in 1997,postgraduate.His main research interests include point cloud representation learning,graph neural network and geometric invariance.
  • Supported by:
    National Key R & D Program(2019YFF0301800),National Natural Science Foundation of China(61379106)and Shandong Provincial Natural Science Foundation(ZR2013FM036,ZR2015FM011).

摘要: 虽然深层神经网络较为成功地解决了点云数据稀疏和不规则等问题,但是点云局部特征的学习仍然是一个非常具有挑战性的问题。现有的用于点云表征学习的网络存在点与点之间相互独立提取特征的问题,为此提出了一种全新的空域图卷积。首先,在构造图结构时提出了自适应空洞K近邻算法,以最大程度地捕获局部拓扑结构信息;其次,在卷积中加入了卷积核每条边与感受野图之间的角度特征,保证了更具鉴别力的特征提取;最后,为了充分利用局部特征,提出了一种全新的图金字塔池化,以更好地融合多尺度特征。在标准公开数据集ModelNet40和ShapeNet上测试该算法,分别取得了93.2%与86.5%的准确度。实验结果表明,该算法在点云表征学习中处于领先水平。

关键词: 表征学习, 点云, 局部特征, 图卷积神经网络

Abstract: Although the sparseness and irregularity of point cloud data have been successfully solved by deep neural networks.However,how to learn the local features of point clouds is still a challenging problem.Existing networks for point cloud representation learning have the problem of extracting features independently between points and points.To this end,a new spatial graph convolution is proposed.Firstly,an adaptive hole K-nearest neighbor algorithm is proposed when constructing the graph structure to maximize local topo-logical structure information.Secondly,the angle feature between each edge of the convolution kernel and the receptive field map is added to the convolution,which ensures more discriminative feature extraction.Finally,in order to make full use of local features,a novel graph pyramid pooling is proposed.This algorithm is tested on the standard public data sets ModelNet40 and ShapeNet,and the accuracy is 93.2% and 86.5% respectively.Experimental results show that the proposed algorithm is at a leading level in point cloud representation learning.

Key words: Graph neural convolutional networks, Local feature, Point clouds, Representation learning

中图分类号: 

  • TP391.41
[1]GUO Y,WANG H,HU Q,et al.Deep Learning for 3D Point Clouds:A Survey[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2020,43(12):4338-4364.
[2]QI C R,SU H,MO K,et al.Pointnet:Deep learning on point sets for 3d classification and segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2017:652-660.
[3]QI C R,YI L,SU H,et al.PointNet++:deep hierarchical feature learning on point sets in a metric space[EB/OL].https://arxiv.org/pdf/1706.02413.pdf.
[4]LI Y,BU R,SUN M,et al.B:PointCNN:Con-volution on X-transformed points[C]//In Conference and Work-shop on Neural Information Processing Systems(NeurIPS).2018:820-830.
[5]LI R,LI X,HENG P A,et al.Pointaugment:an au-to-augmentation framework for point cloud classifica-tion[C]//Proceedings of the IEEE/CVF Conference on Com-puter Vision and Pattern Recognition.2020:6378-6387.
[6]JIE Z A,GC A,SH A,et al.Graph neural networks:A review of methods and applications[J].AI Open,2020,1:57-81.
[7]WANG Y,SUN Y,LIU Z,et al.Dynamic graph cnn for learning on point clouds[J].ACM Transactions On Graphics(tog),2019,38(5):1-12.
[8]LIN Z H,HUANG S Y,WANG Y C F.Convolution in the cloud:Learning deformable kernels in 3D graph convolution net-works for point cloud analysis[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:1800-1809.
[9]ZHAO H,JIANG L,FU C W,et al.PointWeb:Enhancing local neighborhood features for point cloud processing[C]//CVPR.2019:5565-5573.
[10]YANG J,ZHANG Q,NI B,et al.Modeling Point Clouds With Self-Attention and Gumbel Subset Sampling[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).IEEE,2019:3323-3332.
[11]CHOY C B,XU D F,GWAK J Y,et al.3D-R2N2:A Unified Approach for Single and Multi-view 3D Object Reconstruction[C]//Proceedings of the European Conference on Computer Vision(ECCV).2016:628-644.
[12]MATURANA D,SCHERER S.VoxNet:A 3D Convolutional Neural Network for Real-time Object Recognition[C]//Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems(IROS).IEEE,2015:922-928.
[13]MENG H Y,GAO L,LAI Y K,et al.VV-Net:Voxel VAE Net with Group Convolutions for Point Cloud Segmentation[J].ar-Xiv:1811.04337,2018.
[14]ROYNARD X,DESCHAUD J E,GOULETTE F.Classification of Point Cloud Scenes with Multiscale Voxel Deep Network[J].arXiv:1804.03583,2018.
[15]YAN Y,MAO Y,LI B.Second:Sparsely embedded convolu-tional detection[J/OL].Sensors,2018:18(10):3337.https://doi.org/10.3390/s18103337.
[16]TE G,HU W,ZHENG A,et al.Rgcnn:Regularized graph cnn for point cloud segmentation[C]//Proceedings of the 26th ACM international conference on Multimedia.2018:746-754.
[17]ZHANG K,HAO M,WANG J,et al.Linked dynamic graphcnn:Learning on point cloud via linking hierarchical features[J].arXiv:1904.10014,2019.
[18]SIMONOVSKY M,KOMODAKIS N.Dynamic edge-conditioned filters in convolutional neural networks on graphs[C]//CVPR.2017:3693-3702.
[19]LI G,MÜLLER M,THABET A,et al.DeepGCNs:Can GCNs Go as Deep as CNNs? [C]//ICCV.2019:9267-9276.
[20]SHEN Y,FENG C,YANG Y,et al.Mining point cloud localstruc-tures by kernel correlation and graph pooling[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:4548-4557.
[21]XU Q,SUN X,WU C Y,et al.Grid-gcn for fast and scalable point cloud learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:5661-5670.
[22]HU H,WANG F,LE H.VA-GCN:A Vector Attention Graph Convolution Network for learning on Point Clouds[J/OL].https://arxiv.org/pdf/2106.00227.pdf.
[23]HE K,ZHANG X,REN S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEETran-sactions on Pattern Analysis and Machine Intelligence,2015,37(9):1904-1916.
[24]SHILANE P,MIN P,KAZHDAN M,et al.The princeton shape benchmark[C]//Proceedings Shape Modeling Applications,2004.IEEE,2004:167-178.
[25]YI L,KIM V G,CEYLAN D,et al.A scalable active framework for region annotation in 3d shape collections[J].ACM Transactions on Graphics(ToG),2016,35(6):1-12.
[26]KLOKOV R,LEMPITSKY V.Escape from cells:Deep kd-networks for the recognition of 3d point cloud models[C]//2017 IEEE International Conference on Computer Vision(ICCV).2017:863-872.
[27]HAN X F,HE Z Y,CHEN J,et al.Cross-Level Cross-ScaleCross-Attention Network for Point Cloud Representation[EB/OL].https://arxiv.org/pdf/2104.13053.pdf.
[1] 黄丽, 朱焱, 李春平.
基于异构网络表征学习的作者学术行为预测
Author’s Academic Behavior Prediction Based on Heterogeneous Network Representation Learning
计算机科学, 2022, 49(9): 76-82. https://doi.org/10.11896/jsjkx.210900078
[2] 杨文坤, 原晓佩, 陈小锋, 郭睿.
三维激光雷达点云空间多特征分割
Spatial Multi-feature Segmentation of 3D Lidar Point Cloud
计算机科学, 2022, 49(8): 143-149. https://doi.org/10.11896/jsjkx.210300275
[3] 檀莹莹, 王俊丽, 张超波.
基于图卷积神经网络的文本分类方法研究综述
Review of Text Classification Methods Based on Graph Convolutional Network
计算机科学, 2022, 49(8): 205-216. https://doi.org/10.11896/jsjkx.210800064
[4] 彭云聪, 秦小林, 张力戈, 顾勇翔.
面向图像分类的小样本学习算法综述
Survey on Few-shot Learning Algorithms for Image Classification
计算机科学, 2022, 49(5): 1-9. https://doi.org/10.11896/jsjkx.210500128
[5] 封雷, 朱登明, 李兆歆, 王兆其.
一种基于遮罩的稀疏点云滤波算法
Sparse Point Cloud Filtering Algorithm Based on Mask
计算机科学, 2022, 49(5): 25-32. https://doi.org/10.11896/jsjkx.210600129
[6] 李子仪, 周夏冰, 王中卿, 张民.
基于用户关联的立场检测
Stance Detection Based on User Connection
计算机科学, 2022, 49(5): 221-226. https://doi.org/10.11896/jsjkx.210400135
[7] 高越, 傅湘玲, 欧阳天雄, 陈松龄, 闫晨巍.
基于时空自适应图卷积神经网络的脑电信号情绪识别
EEG Emotion Recognition Based on Spatiotemporal Self-Adaptive Graph ConvolutionalNeural Network
计算机科学, 2022, 49(4): 30-36. https://doi.org/10.11896/jsjkx.210900200
[8] 李浩, 张兰, 杨兵, 杨海潇, 寇勇奇, 王飞, 康雁.
融合双重权重机制和图卷积神经网络的微博细粒度情感分类
Fine-grained Sentiment Classification of Chinese Microblogs Combining Dual Weight Mechanismand Graph Convolutional Neural Network
计算机科学, 2022, 49(3): 246-254. https://doi.org/10.11896/jsjkx.201200073
[9] 苗启广, 辛文天, 刘如意, 谢琨, 王泉, 杨宗凯.
面向智慧教育行为分析的图卷积骨架动作识别方法
Graph Convolutional Skeleton-based Action Recognition Method for Intelligent Behavior Analysis
计算机科学, 2022, 49(2): 156-161. https://doi.org/10.11896/jsjkx.220100061
[10] 张虎, 柏萍.
融入句子中远距离词语依赖的图卷积短文本分类方法
Graph Convolutional Networks with Long-distance Words Dependency in Sentences for Short Text Classification
计算机科学, 2022, 49(2): 279-284. https://doi.org/10.11896/jsjkx.201200062
[11] 张玮琪, 汤轶丰, 李林燕, 胡伏原.
基于场景图的段落生成序列图像方法
Image Stream From Paragraph Method Based on Scene Graph
计算机科学, 2022, 49(1): 233-240. https://doi.org/10.11896/jsjkx.201100207
[12] 叶洪良, 朱皖宁, 洪蕾.
基于CQT和梅尔频谱的带有人声的音乐风格转换方法
Music Style Transfer Method with Human Voice Based on CQT and Mel-spectrum
计算机科学, 2021, 48(6A): 326-330. https://doi.org/10.11896/jsjkx.200900104
[13] 梁浩宏, 古天龙, 宾辰忠, 常亮.
联合学习用户端和项目端知识图谱的个性化推荐
Combining User-end and Item-end Knowledge Graph Learning for Personalized Recommendation
计算机科学, 2021, 48(5): 109-116. https://doi.org/10.11896/jsjkx.200600115
[14] 赵新灿, 常寒星, 金仁标.
3D点云形状补全GAN
3D Point Cloud Shape Completion GAN
计算机科学, 2021, 48(4): 192-196. https://doi.org/10.11896/jsjkx.200100048
[15] 王省, 康昭.
基于光滑表示的半监督分类算法
Smooth Representation-based Semi-supervised Classification
计算机科学, 2021, 48(3): 124-129. https://doi.org/10.11896/jsjkx.200700078
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!