Laser & Optoelectronics Progress, Volume. 57, Issue 10, 101510(2020)

Point Cloud Model Recognition and Classification Based on K-Nearest Neighbor Convolutional Neural Network

Ting Yu and Jun Yang*
Author Affiliations
  • School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou, Gansu 730070, China
  • show less
    Figures & Tables(11)
    Construction process of the local neighborhood
    Schematic diagram of feature extraction process
    JANet model
    KNN-CNN model
    Graph of activation function
    Three representations of three-dimensional models. (a) Rendered model; (b) mesh model; (c) 3D point cloud model
    Examples of misclassified point cloud models. (a) Television stand; (b) chair; (c) plant; (d) flower pot
    • Table 1. Recognition accuracy of different algorithms on ModelNet40 dataset

      View table

      Table 1. Recognition accuracy of different algorithms on ModelNet40 dataset

      MethodInputMA /%OA /%
      PointNet[4]102486.289.2
      PointNet++[20]1024-90.7
      PointNet++[20]normal-91.9
      KCNet[22]1024-91.0
      Kd-Net(depth 10)[24]102486.390.6
      Kd-Net(depth 15)[24]3276888.591.8
      Ours102489.992.0
    • Table 2. Comparison of recognition accuracy of each category model in ModelNet40 test setunit:%

      View table

      Table 2. Comparison of recognition accuracy of each category model in ModelNet40 test setunit:%

      CategoryPointNet[4]Our model
      Airplane100.0100.0
      Bathtub86.094.0
      Bed97.099.0
      Bench70.075.0
      Bookshelf91.099.0
      Bottle94.098.0
      Bow90.0100.0
      Car98.099.0
      Chair97.098.0
      Cone95.0100.0
      Cup80.075.0
      Curtain90.095.0
      Desk80.287.2
      Door80.095.0
      Dresser72.176.7
      Flower pot20.020.0
      Glass box98.092.0
      Guitar100.0100.0
      Keyboard100.0100.0
      Lamp95.090.0
      Laptop100.0100.0
      Mantel95.096.0
      Monitor96.099.0
      Night stand72.182.6
      Person90.0100.0
      Piano86.095.0
      Plant76.082.0
      Radio75.090.0
      Range hood92.095.0
      Sink80.080.0
      Sofa95.096.0
      Stairs85.095.0
      Stool85.085.0
      Table81.081.0
      Tent95.095.0
      Toilet98.098.0
      Television stand79.091.0
      Vase80.083.0
      Wardrobe60.070.0
      Xbox85.090.0
    • Table 3. Effect of k value on OA

      View table

      Table 3. Effect of k value on OA

      k1k2OA /%Time /h
      151588.572.0
      201091.272.1
      202091.393.3
      302091.6107.3
    • Table 4. Recognition accuracy of different optimization schemesunit:%

      View table

      Table 4. Recognition accuracy of different optimization schemesunit:%

      Activate functionOA
      SELU91.3
      ReLU+BN91.6
      ELU+BN91.8
      SELU+BN92.0
    Tools

    Get Citation

    Copy Citation Text

    Ting Yu, Jun Yang. Point Cloud Model Recognition and Classification Based on K-Nearest Neighbor Convolutional Neural Network[J]. Laser & Optoelectronics Progress, 2020, 57(10): 101510

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Machine Vision

    Received: Sep. 9, 2019

    Accepted: Sep. 17, 2019

    Published Online: May. 8, 2020

    The Author Email: Jun Yang (yangj@mail.lzjtu.cn)

    DOI:10.3788/LOP57.101510

    Topics