首页 > 论文 > 激光与光电子学进展 > 56卷 > 21期(pp:211004--1)

基于图卷积网络的深度学习点云分类模型

Deep Learning Model for Point Cloud Classification Based on Graph Convolutional Network

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

PointNet是三维点云分类中具有代表性的研究成果,该模型开创性地利用深度学习模型对点云进行分类,取得了较好的效果。但是PointNet模型只考虑点云的全局特征而忽略每个点的局部信息,为弥补这个缺陷,提出基于图卷积网络的点云分类模型。在PointNet模型中插入一个kNN graph层,通过在点云空间构造k近邻图,利用图结构有效地获取点云的局部信息,从而提高整体点云分类准确率。分类实验在ModelNet40数据集上进行,对比不同近邻值k对输出精度的影响,结果表明在k取20时,分类准确率最高,达到了93.2%,比PointNet高4.0%。

Abstract

PointNet is one of the representative research results obtained from three-dimensional point cloud classification, which innovatively employs a deep learning model for point cloud classification and achieves good results. However, PointNet does not capture local information of each point, and it considers only the global features of point clouds. Herein, we propose a model for point cloud classification based on graph convolutional networks to solve this problem, in which a k-nearest neighbor (kNN) graph layer is designed and plugged into a PointNet model. The local information of point clouds can be effectively obtained by constructing the kNN graph layer in the point cloud space, which can improve the accuracy of point cloud classification. The point cloud classification experiment is conducted on the ModelNet40 dataset, and the effects of the different neighbor values of k on the output accuracy are compared. The results demonstrate that the highest classification accuracy is achieved when k is 20, reaching 93.2%, which is 4.0% higher than that of PointNet.

广告组6 - 调制器
补充资料

中图分类号:TP391

DOI:10.3788/LOP56.211004

所属栏目:图像处理

基金项目:天津市科技特派员项目;

收稿日期:2019-04-01

修改稿日期:2019-04-30

网络出版日期:2019-11-01

作者单位    点击查看

王旭娇:河北工业大学电子信息工程学院, 天津 300401
马杰:河北工业大学电子信息工程学院, 天津 300401
王楠楠:河北工业大学电子信息工程学院, 天津 300401
马鹏飞:河北工业大学电子信息工程学院, 天津 300401
杨立闯:河北工业大学电子信息工程学院, 天津 300401

联系人作者:马杰(13163152009@163.com)

备注:天津市科技特派员项目;

【1】Lei Z, Xi X H, Wang C et al. Building point clouds extraction from airborne LiDAR data based on decision tree method. Laser & Optoelectronics Progress. 55(8), (2018).
雷钊, 习晓环, 王成 等. 决策树约束的建筑点云提取方法. 激光与光电子学进展. 55(8), (2018).

【2】Tong G F, Du X C, Li Y et al. Three-dimensional point cloud classification of large outdoor scenes based on vertical slice sampling and centroid distance histogram. Chinese Journal of Lasers. 45(10), (2018).
佟国峰, 杜宪策, 李勇 等. 基于切片采样和质心距直方图特征的室外大场景三维点云分类. 中国激光. 45(10), (2018).

【3】Jia Y J, Xiong F G, Han X et al. Multi-scale keypoint detection based on SHOT. Laser & Optoelectronics Progress. 55(7), (2018).
贾勇杰, 熊风光, 韩燮 等. 基于SHOT的多尺度关键点检测技术研究. 激光与光电子学进展. 55(7), (2018).

【4】Charles R Q, Su H, Kaichun M et al. PointNet: deep learning on point sets for 3D classification and segmentation. [C]∥2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 21-26, 2017, Honolulu, HI, USA. New York: IEEE. 652-660(2017).

【5】Li Y G, Yu R, Shahabi C et al. -02-22)[2019-2-20]. https:∥arxiv. org/abs/1707, (2018).

【6】Yu B and Yin H T. -07-12)[2019-2-20]. https:∥arxiv. org/abs/1709, (2018).

【7】Yan S J, Xiong Y J and Lin D H. Spatial temporal graph convolutional networks for skeleton-based action recognition. [C]∥The Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18), February 2-7, 2018, Hilton New Orleans Riverside, New Orleans, Louisiana, USA. New York: AAAI. 7444-7452(2018).

【8】Jain A, Zamir A R, Savarese S et al. Structural-RNN: deep learning on spatio-temporal graphs. [C]∥2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 27-30, 2016, Las Vegas, NV, USA. New York: IEEE. 5308-5317(2016).

【9】Pan S R, Wu J, Zhu X Q et al. Joint structure feature exploration and regularization for multi-task graph classification. IEEE Transactions on Knowledge and Data Engineering. 28(3), 715-728(2016).

【10】Cignoni P, Callieri M, Corsini M et al. Meshlab: an open-source mesh processing tool. Eurographics Italian Chapter Conference. 2008, 129-136(2008).

【11】Wu Z R, Song S R, Khosla A et al. 3D ShapeNets: a deep representation for volumetric shapes. [C]∥2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 7-12, 2015, Boston, MA, USA. New York: IEEE. 1912-1920(2015).

【12】Maturana D and Scherer S. VoxNet: a 3D convolutional neural network for real-time object recognition. [C]∥2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), September 28-October 2, 2015, Hamburg, Germany. New York: IEEE. 922-928(2015).

【13】Qi C R, Yi L, Su H et al. Pointnet++: deep hierarchical feature learning on point sets in a metric space. [C]∥Advances in Neural Information Processing Systems, December 4-9, 2017, Long Beach Convention Center, Long Beach. New York: NIPS. 5099-5108(2017).

【14】Klokov R and 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), October 22-29, 2017, Venice, Italy. New York: IEEE. 863-872(2017).

引用该论文

Wang Xujiao,Ma Jie,Wang Nannan,Ma Pengfei,Yang Lichaung. Deep Learning Model for Point Cloud Classification Based on Graph Convolutional Network[J]. Laser & Optoelectronics Progress, 2019, 56(21): 211004

王旭娇,马杰,王楠楠,马鹏飞,杨立闯. 基于图卷积网络的深度学习点云分类模型[J]. 激光与光电子学进展, 2019, 56(21): 211004

被引情况

【1】张佳颖,赵晓丽,陈正. 基于深度学习的点云语义分割综述. 激光与光电子学进展, 2020, 57(4): 40002--1

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF