2.624

2020影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

RFNet: 用于三维点云分类的卷积神经网络

单铉洋 孙战里 曾志刚

单铉洋, 孙战里, 曾志刚. RFNet: 用于三维点云分类的卷积神经网络. 自动化学报, 2022, 48(x): 1−10 doi: 10.16383/j.aas.c210532
引用本文: 单铉洋, 孙战里, 曾志刚. RFNet: 用于三维点云分类的卷积神经网络. 自动化学报, 2022, 48(x): 1−10 doi: 10.16383/j.aas.c210532
Shan Xuan-Yang, Sun Zhan-Li, Zeng Zhi-Gang. RFNet: convolutional neural network for 3D point cloud classification. Acta Automatica Sinica, 2022, 48(x): 1−10 doi: 10.16383/j.aas.c210532
Citation: Shan Xuan-Yang, Sun Zhan-Li, Zeng Zhi-Gang. RFNet: convolutional neural network for 3D point cloud classification. Acta Automatica Sinica, 2022, 48(x): 1−10 doi: 10.16383/j.aas.c210532

RFNet: 用于三维点云分类的卷积神经网络

doi: 10.16383/j.aas.c210532
基金项目: 国家自然科学基金(61972002)资助
详细信息
    作者简介:

    单铉洋:安徽大学电气工程与自动化学院硕士研究生.主要研究方向为模式识别, 计算机视觉, 深度学习.E-mail: shanxy128@163.com

    孙战里:安徽大学人工智能学院教授.主要研究方向为模式识别, 机器学习, 图像信号处理. 本文通信作者.E-mail: zhlsun2006@126.com

    曾志刚:华中科技大学人工智能与自动化学院教授.主要研究方向为神经网络, 智能计算, 模式识别.E-mail: zgzeng@hust.edu.cn

RFNet: Convolutional neural network for 3D point cloud classification

Funds: Supported by National Natural Science Foundation of P. R. China (61972002)
More Information
    Author Bio:

    SHAN Xuan-Yang Master student at the School of Electrical Engineering and Automation, Anhui University. His research interest covers pattern recognition, computer vision, and deep learning

    SUN Zhan-Li Professor at the School of Artificial Intelligence, Anhui University. His research interests include machine learning, and image and signal processing. Corresponding author of this paper

    ZENG Zhi-Gang Professor at the School of Artificial Intelligence and Automation, Huazhong University of Science and Technology. His research interests include neural networks, computational intelligence and pattern recognition

  • 摘要: 由于点云的非结构性和无序性, 目前已有的点云分类网络在精度上仍然需要进一步提高. 通过考虑局部结构的构建、全局特征聚合和损失函数改进三个方面, 本文构造了一个有效的点云分类网络. 首先, 针对点云的非结构性,通过学习中心点特征与近邻点特征之间的关系, 为不规则的近邻点分配不同的权重, 以此构建局部结构. 此外,使用注意力的思想, 提出了加权平均池化, 通过自注意力的方式, 学习每个高维特征的注意力分数, 在应对点云无序性的同时, 可以有效地聚合冗余的高维特征. 另外,利用了交叉熵损失与中心损失之间的互补关系, 提出了联合损失, 在增大类间距离的同时减小了类内距离, 进一步提高了网络的分类能力. 本文在合成数据集ModelNet40、ShapeNetCore和真实世界数据集ScanObjectNN上进行了实验, 与目前性能最好的多个网络相比较, 验证了本文整体网络结构的优越性.
  • 图  1  整体网络结构图

    Fig.  1  Diagram of overall network structure

    图  2  中心点和近邻点的局部结构示意图

    Fig.  2  Diagram of local structure between the central point and its neighbors

    图  3  局部特征提取结构图

    Fig.  3  Diagram of local feature extraction structure

    图  4  加权平均池化结构图

    Fig.  4  Diagram of weighted average pooling structure

    图  5  联合损失

    Fig.  5  Joint loss

    图  6  混淆矩阵

    Fig.  6  Confusion matrix

    图  7  采样密度实验结果对比

    Fig.  7  Comparison of experiment results with different sampling densities

    图  8  添加高斯噪声前后点云样本的对照图

    Fig.  8  A comparison of one sample with and without Gaussian noise

    表  1  实验配置

    Table  1  Experimental configuration

    项目详情
    操作系统CentOS Linux 7
    GPUTesla V100 32G
    CUDA10.0
    CUDNN7.0
    Python3.6.0
    Pytorch1.3.0
    下载: 导出CSV

    表  2  在ModelNet40数据集上的实验结果

    Table  2  Experimental results on ModelNet40

    MethodInputmAcc(%)OA(%)
    PointNet [3]1k86.089.2
    PointNet++[10]1k-90.7
    Spec-GCN[11]1k-91.5
    DGCNN[12]1k90.292.9
    RSCNN[13]1k-92.9
    PCT[16]1k-93.2
    ECC[21]1k83.287.4
    RMFP-DNN[22]1k88.992.6
    PointCNN[23]1k88.192.2
    P2Sequence[24]1k90.492.6
    Point Transformer [25]1k-92.8
    Octant-CNN[26]1k88.791.9
    DRNet[27]1k-93.1
    AdaptConv[28]1k90.793.4
    RFNet1k91.293.6
    PointNet++[10]5k+nor-91.9
    DGCNN[12]2k90.793.5
    RSCNN[13]1k+voting-93.6
    SpiderCNN[29]5k+nor-92.4
    RFNet2k91.494.0
    下载: 导出CSV

    表  3  不同网络的易错模型分类对比

    Table  3  Classification comparison of error-prone models for different networks

    易错模型真实标签Point- Net[3]DG- CNN[12]RFNet
    植物植物植物植物
    花盆植物植物植物
    花瓶花盆花瓶花瓶
    花瓶瓶子花瓶花瓶
    杯子花盆花盆杯子
    下载: 导出CSV

    表  4  在ShapeNetCore数据集上的实验结果

    Table  4  Experimental results on ShapeNetCore

    MethodInputOA(%)
    PointNet[3]1k83.7
    PointNet++[10]1k85.1
    DGCNN[12]1k84.7
    P2sequence[24]1k85.2
    SpiderCNN[29]1k+nor85.3
    KPConv[30]1k86.2
    RFNet1k88.3
    下载: 导出CSV

    表  5  在ScanObjectNN数据集上的实验结果

    Table  5  Experimental results on ScanObjectNN

    MethodmAcc (%)OA (%)BagBinBoxSofaDeskShelftableDoorBedCabinetChairDisplayPillowSinkToilet
    PointNet[3] 63.4 68.2 36.1 69.8 10.5 76.7 50.0 72.6 67.8 93.8 61.8 62.6 89.0 73.0 67.6 64.2 55.3
    PointNet++[10] 75.4 77.9 49.4 84.4 31.6 90.5 74.0 72.6 72.6 85.2 75.5 77.4 91.3 79.4 81.0 80.8 85.9
    DGCNN[12] 73.6 78.1 49.4 82.4 33.1 91.4 63.3 79.3 77.4 89.0 64.5 83.9 91.8 77.0 77.1 75.0 69.4
    PointCNN[23] 75.1 78.5 57.8 82.9 33.1 91.9 65.3 84.2 67.4 84.8 80.0 83.6 92.6 78.4 80.0 72.5 71.8
    SpiderCNN [29] 69.8 73.7 43.4 75.9 12.8 90.5 65.3 78.0 65.9 91.4 69.1 74.2 89.0 74.5 80.0 65.8 70.6
    RFNet 76.3 79.6 54.0 81.9 34.1 92.2 74.1 79.9 72.1 91.5 81.3 83.5 91.3 80.2 82.6 78.9 67.0
    下载: 导出CSV

    表  6  不同模块的消融研究

    Table  6  Ablation studies about different modules

    MethodRFConvWAPJLmAcc (%)OA (%)
    0×××90.292.9
    1××91.093.3
    2×91.193.5
    3×91.193.4
    491.293.6
    下载: 导出CSV

    表  7  高斯噪声鲁棒性实验

    Table  7  Robustness experiment of Gaussian noise

    Method无噪音有噪音
    mAcc (%)OA (%)mAcc (%)OA (%)
    PointNet[3]86.089.283.487.6
    PointNet++[10]-90.7-89.6
    DGCNN[12]90.292.989.792.6
    RSCNN[13]-92.9-92.3
    RFNet91.293.691.093.5
    下载: 导出CSV

    表  8  不同特征关系的消融实验

    Table  8  Ablation studies about different relationship between features

    Method${e_{ij}}$$di{s_{ij}}$${l_{ij}}$${s_{ij}}$OA (%)
    1×××93.3
    2××93.5
    3×93.6
    4××93.1
    5×93.5
    693.2
    下载: 导出CSV

    表  9  网络复杂度对比

    Table  9  Comparison of network complexity

    MethodParams (M)FLOPs(G)OA (%)
    PointNet[3]3.470.4589.2
    PointNet++[10]1.744.0991.9
    DGCNN[12]1.812.4392.9
    PCT[16]2.882.3293.2
    KPConv[30]14.3-92.9
    RFNet2.362.9593.6
    下载: 导出CSV
  • [1] Guo Y, Bennamoun M, Sohel F, Min L, Wan J. 3D object recognition in cluttered scenes with local surface features: a survey. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 36(11): 2270-2287
    [2] 田永林, 沈宇, 李强, 王飞跃. 平行点云: 虚实互动的点云生成与三维模型进化方法. 自动化学报, 2020, 46(12): 2572-2582

    Tian Yong-Lin, Shen Yu, Li Qiang, Wang Fei-Yue. Parallel point clouds: point clouds generation and 3d model evolution via virtual-real interaction. Acta Automatica Sinica, 2020, 46(12): 2572-2582
    [3] Qi C R, Su H, Mo K, Guibas L J. Pointnet: Deep learning on point sets for 3D classification and segmentation. In: Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI: IEEE, 2017. 77−85
    [4] Guo Y, Wang H, Hu Q, Liu H, Bennamoun M. Deep learning for 3d point clouds: a survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020. 43(12): 4438-4364
    [5] Su H, Maji S, Kalogerakis E, Learned-Miller E. Multi-view convolutional neural networks for 3D shape recognition. In: Proceedings of the 2015 IEEE International Conference on Computer Vision (ICCV). Santiago, Chile: IEEE, 2015. 945−953
    [6] Yang Z, Wang L. Learning relationships for multi-view 3d object recognition. In: Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Seoul, South Korea: IEEE, 2019. 7504-7513
    [7] 徐思雨, 祝继华, 田智强, 李垚辰, 庞善民. 逐步求精的多视角点云配准方法. 自动化学报, 2019, 45(8): 1486-1494

    Xu Si-Yu, Zhu Ji-Hua, Tian Zhi-Qiang, Li Yao-Chen, Pang Shan-Min. Stepwise refinement approach for registration of multi-view point sets. Acta Automatica Sinica, 2019, 45(8): 1486-1494
    [8] Maturana D, Scherer S. Voxnet: a 3d convolutional neural network for real-time object recognition. In: Proceedings of the 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Hamburg, Germany: IEEE, 2015. 922-928
    [9] Riegler G, Osman Ulusoy A, Geiger A. OctNet: Learning deep 3D representations at high resolutions. In: Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI: IEEE, 2017. 6620−6629
    [10] Qi C R, Yi L, Su H, Guibas L J. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. In: Proceedings of the 2017 Advances in Neural Information Processing Systems. Long Beach, CA, 2017. 5099−5108
    [11] Wang C, Samari B, Siddiqi K. Local spectral graph convolution for point set feature learning. In: Proceedings of the 2018 European Conference on Computer Vision (ECCV). Munich, Germany, 2018. 56-71
    [12] Wang Y, Sun Y, Liu Z, Sarma S E, Bronstein M M, Solomon J M. Dynamic graph cnn for learning on point clouds. ACM Transactions on Graphics, 2019, 38(5): 1-12
    [13] Liu Y, Fan B, Xiang S, Pan C. Relation-shape convolutional neural network for point cloud analysis. In: Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, CA, USA: IEEE, 2019.8895-8904
    [14] Wang L, Huang Y, Hou Y, Zhang S, Shan J. Graph attention convolution for point cloud semantic segmentation. In: Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, CA: IEEE, 2019.10288-10297
    [15] Chen C, Fragonara L Z, Tsourdos A. Gapointnet: graph attention based point neural network for exploiting local feature of point cloud. Neurocomputing, 2021, 438: 122-132 doi: 10.1016/j.neucom.2021.01.095
    [16] Guo M H, Cai J X, Liu Z N, Mu T J, Martin R R, Hu S M. Pct: point cloud transformer. Computational Visual Media, 2021, 7(2): 187-299 doi: 10.1007/s41095-021-0229-5
    [17] Wen Y, Zhang K, Li Z, Qiao Y. A discriminative feature learning approach for deep face recognition. In: Proceedings of the 2016 European Conference on Computer Vision. Amsterdam, Netherlands, 2016. 499-515
    [18] Wu Z, Song S, Khosla A, Yu F, Zhang L, Tang X, et al. 3D shapenets: a deep representation for volumetric shapes. In: Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, MA: IEEE, 2015.1912-1920
    [19] Yi L, Kim V G, Ceylan D, Shen I C, Yan M, Su H, et al. A scalable active framework for region annotation in 3d shape collections. ACM Transactions on Graphics, 2016, 35(6): 1-12
    [20] Uy M A, Pham Q H, Hua B S, Nguyen T, Yeung S K. Revisiting point cloud classification: a new benchmark dataset and classification model on real-world data. In: Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, South Korea: IEEE, 2019.1588-1597
    [21] Simonovsky M, Komodakis N. Dynamic edge-conditioned filters in convolutional neural networks on graphs. In: Proceedings of the 2017 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI: IEEE, 2017.3693-3702
    [22] 田钰杰, 管有庆, 龚锐. 一种鲁棒的多特征点云分类分割深度神经网络. 计算机工程, 2021, 47(11): 234-240

    Tian Yu-Jie, Guan You-Qing, Gong Rui. A robust deep neural network for multi-feature point cloud classification and segmentation. Computer Engineering, 2021, 47(11): 234-240
    [23] Li Y, Bu R, Sun M, Chen B. Pointcnn: convolution on x-transformed points. In: Proceeding of the 2018 Advances in Neural Information Processing System. Montreal, Canada, 2018. 828-838
    [24] Liu X, Han Z, Liu Y, Zwicker M. Point2sequence: learning the shape representation of 3d point clouds with an attention-based sequence to sequence network. In: Proceedings of the 2019 AAAI Conference on Artificial Intelligence. Honolulu, HI: AAAI, 2019.8778-8785
    [25] Engel N, Belagiannis V, Dietmayer K. Point transformer. IEEE Access, 2021, 9: 134826-134840 doi: 10.1109/ACCESS.2021.3116304
    [26] 许翔, 帅惠, 刘青山. 基于卦限卷积神经网络的3D点云分析. 自动化学报, 2021, 47(12): 2791−2800

    Xu Xiang, Shuai Hui, Liu Qing-Shan. Octant convolutional neural network for 3D point cloud analysis. Acta Automatica Sinica, 2021, 47(12): 2791−2800
    [27] Qiu S, Anwar S, Barnes N. Dense-resolution network for point cloud classification and segmentation. In: Proceedings of the 2021 IEEE Winter Conference on Applications of Computer Vision (WACV). Electr Network: IEEE, 2021.3812-3821
    [28] Zhou H, Feng Y, Fang M, Wei M, Qin J, Lu T. Adaptive graph convolution for point cloud analysis. In: Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, Canada: IEEE, 2021. 4965-4974
    [29] Xu Y, Fan T, Xu M, Long Z, Yu Q. Spidercnn: deep learning on point sets with parameterized convolutional filters. In: Proceedings of the 2018 European Conference on Computer Vision. Munich, Germany, 2018.90-105
    [30] Thomas H, Qi C R, Deschaud J E, Marcotegui B, Goulette F, Guibas L. Kpconv: flexible and deformable convolution for point clouds. In: Proceedings of the 2020 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, South Korea: IEEE, 2020.6420-6429
  • 加载中
计量
  • 文章访问数:  332
  • HTML全文浏览量:  132
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-06-15
  • 录用日期:  2022-02-10
  • 网络出版日期:  2022-05-05

目录

    /

    返回文章
    返回