Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 60, Issue 4, 0410020(2023)

Point-Cloud Instance Segmentation-Based Robust Multi-Target Pose Estimation

Yaohua Liu1,1,2,2,3,3,4、">">">*, Yue Ma1,1,2,2,4,4、">">">, and Min Xu1,1,2,2,4,4、">">">
Author Affiliations
  • 1Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, Liaoning, China
  • 2Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang 110169, Liaoning, China
  • 3University of Chinese Academy of Sciences, Beijing 100049, China
  • 4Key Laboratory on Intelligent Detection and Equipment Technology of Liaoning Province, Shenyang 100179, Liaoning, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (14)
    Equations (10)
    References (24)
    Tools
    Paper Information
    Topics
    Figures & Tables(14)
    Multi-target feature matching with similar local geometry

    Fig. 1. Multi-target feature matching with similar local geometry

    Download full size
    Framework of proposed algorithm

    Fig. 2. Framework of proposed algorithm

    Download full size
    Point cloud local feature aggregation and classification

    Fig. 3. Point cloud local feature aggregation and classification

    Download full size
    CV-Lab 3D sample imbalance in model classification. (a) Training classifier without feature sampling; (b) training classifier after feature sampling

    Fig. 4. CV-Lab 3D sample imbalance in model classification. (a) Training classifier without feature sampling; (b) training classifier after feature sampling

    Download full size
    Point-to-plane ICP diagram

    Fig. 5. Point-to-plane ICP diagram

    Download full size
    Point cloud models in CV-Lab 3D dataset 1. (a) Dragon_vrip; (b) Armadillo; (c) Bunny;（d） Happy_vrip；（e） Xyzrgb_dragon；（f） Statuette

    Fig. 6. Point cloud models in CV-Lab 3D dataset 1. (a) Dragon_vrip; (b) Armadillo; (c) Bunny;（d） Happy_vrip；（e） Xyzrgb_dragon；（f） Statuette

    Download full size
    Point cloud models in UWA dataset. (a) T-rex; (b) Chef;（c） Chicken；（d） Parasaurolophus；（e） Rhino

    Fig. 7. Point cloud models in UWA dataset. (a) T-rex; (b) Chef;（c） Chicken；（d） Parasaurolophus；（e） Rhino

    Download full size
    CV-Lab 3D dataset 1 scene 17. (a) Models in scene; (b)-(e) local feature matching before optimization; (f) initial pose estimation result before optimization; (g) point-to-plane ICP pose refine result before optimization; (h)-(k) local feature matching after optimization using proposed algorithm; (l) initial pose estimation result after optimization using proposed algorithm; (m) point-to-plane ICP pose refinement result after optimization using proposed algorithm

    Fig. 8. CV-Lab 3D dataset 1 scene 17. (a) Models in scene; (b)-(e) local feature matching before optimization; (f) initial pose estimation result before optimization; (g) point-to-plane ICP pose refine result before optimization; (h)-(k) local feature matching after optimization using proposed algorithm; (l) initial pose estimation result after optimization using proposed algorithm; (m) point-to-plane ICP pose refinement result after optimization using proposed algorithm

    Download full size
    UWA scene 30. (a) Models in scene; (b)-(e) local feature matching before optimization; (f) initial pose estimation result before optimization; (g) point-to-plane ICP pose refine result before optimization; (h)-(k) local feature matching after optimization using proposed algorithm; (l) initial pose estimation result after optimization using proposed algorithm; (m) point-to-plane ICP pose refine result after optimization using proposed algorithm

    Fig. 9. UWA scene 30. (a) Models in scene; (b)-(e) local feature matching before optimization; (f) initial pose estimation result before optimization; (g) point-to-plane ICP pose refine result before optimization; (h)-(k) local feature matching after optimization using proposed algorithm; (l) initial pose estimation result after optimization using proposed algorithm; (m) point-to-plane ICP pose refine result after optimization using proposed algorithm

    Download full size

    • Table 1. Parameters used in CV-Lab 3D dataset 1

      View table

      Table 1. Parameters used in CV-Lab 3D dataset 1

      AlgorithmParameterValue
      Down sampleVoxel size0.006 m
      DBSCANSearch radius0.008 m
      Min sample6
      RFN estimators100
      Max depth10
      FLANNtNNSR0.85
      RANSACThreshold0.13

    • Table 2. Parameters used in UWA dataset

      View table

      Table 2. Parameters used in UWA dataset

      AlgorithmParameterValue
      Down sampleVoxel size0.40 m
      DBSCANSearch radius0.55 m
      Min sample5
      RFN estimators100
      Max depth10
      FLANNtNNSR0.85
      RANSACThreshold12

    • Table 3. Average accuracy of point cloud local feature matching on CV-Lab 3D dataset 1

      View table

      Table 3. Average accuracy of point cloud local feature matching on CV-Lab 3D dataset 1

      ModelBaselineProposed
      Armadillo0.5710.847
      Bunny0.5630.690
      Dragon_vrip0.5170.679
      Happy_vrip0.6930.934
      Dragon_vr30.4440.816
      Statuette0.6900.804

    • Table 4. Average accuracy of point cloud local feature matching on UWA dataset

      View table

      Table 4. Average accuracy of point cloud local feature matching on UWA dataset

      ModelBaselineProposed
      T-rex0.1680.293
      Chef0.2570.370
      Chicken0.2360.556
      Parasaurolophus0.1940.618

    • Table 5. Pose estimation precision on UWA dataset

      View table

      Table 5. Pose estimation precision on UWA dataset

      ModelBaselineProposed
      Armadillo0.9091.000
      Bunny0.8520.964
      Dragon_vrip0.8440.853
      Happy_vrip0.7500.893
      Dragon_vr30.7860.957
      Statuette0.8800.808
    Tools

    Get Citation

    Copy Citation Text

    Yaohua Liu, Yue Ma, Min Xu. Point-Cloud Instance Segmentation-Based Robust Multi-Target Pose Estimation[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0410020

    Download Citation

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

    Category: Image Processing

    Received: Jan. 20, 2022

    Accepted: Mar. 24, 2022

    Published Online: Feb. 14, 2023

    The Author Email: Liu Yaohua (liuyaohua19@mails.ucas.ac.cn)

    DOI:10.3788/LOP220586

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology