Laser & Optoelectronics Progress, Volume. 61, Issue 18, 1815002(2024)

Two-Step Relocalization Method for Laser Point Clouds Based on Semantic Graph and Semantic Scan Context

Xiaohong Huang, Yuhui Peng*, and Wei Huang
Author Affiliations
  • College of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350116, Fujian, China
  • show less
    Figures & Tables(11)
    Flowchart of two-step relocalization
    Illustration of semantic segmentation for point clouds
    Semantic graph representation. (a) Semantic segmentation; (b) instance segmentation; (c) (d) removing moving and movable objects
    Network structure of node embedding.
    Example of generating semantic scan contest
    Precision-recall curves on KITTI dataset. (a) 00; (b) 02; (c) 05; (d) 06; (e) 07; (f) 08
    Schematic of robustness test
    • Table 1. Maximum F1 values on KITTI dataset

      View table

      Table 1. Maximum F1 values on KITTI dataset

      Method000205060708Mean
      SC0.7500.7820.8950.9680.6620.6070.777
      ISC0.6570.7050.7710.8420.6360.4080.670
      M2DP0.7080.7170.6020.7870.5600.0730.575
      PV0.7790.72705410.8520.6310.0370.595
      ON0.8690.8270.9240.9300.8180.3740.790
      SGPR0.8200.7510.7510.6550.8680.7500.766
      Ours-SK0.9290.8700.9210.9740.8640.8610.903
      Ours-RN0.9410.8720.9390.9730.8690.9250.920
    • Table 2. Maximum F1 values on KITTI dataset when point clouds are randomly occluded with 30°

      View table

      Table 2. Maximum F1 values on KITTI dataset when point clouds are randomly occluded with 30°

      Method000205060708MeanCmp
      M2DP0.1230.0640.1540.5600.2480.0060.193-0.381
      SC0.7290.7710.8650.9660.5900.5170.783-0.038
      SGPR0.7950.7090.7160.6430.8120.7170.732-0.033
      Ours-SK0.9270.8320.8830.9410.8160.8270.871-0.032
      Ours-RN0.9370.8410.8840.9540.8350.8560.885-0.035
    • Table 3. Maximum F1 values on KITTI dataset when point clouds are randomly rotated around z-axis

      View table

      Table 3. Maximum F1 values on KITTI dataset when point clouds are randomly rotated around z-axis

      Method000205060708MeanCmp
      M2DP0.2970.2800.2450.5590.2890.2520.320-0.254
      SC0.7500.7830.8940.9680.6760.6090.780+0.003
      SGPR0.8190.7500.7480.6500.8610.7460.762-0.003
      Ours-SK0.9420.8570.9270.9890.9410.8090.910+0.007
      Ours-RN0.9500.8600.9490.9900.9450.8140.918-0.002
    • Table 4. Efficiency evaluation

      View table

      Table 4. Efficiency evaluation

      MethodSegmentationDescriptionRetrievalICPTotal
      RoughExactRoughExact
      SC235.32528.415263.740
      ISC3.0940.8003.894
      SSC2.5630.0662.1264.755
      Ours83.33350.0412.5630.0710.0662.126138.200
    Tools

    Get Citation

    Copy Citation Text

    Xiaohong Huang, Yuhui Peng, Wei Huang. Two-Step Relocalization Method for Laser Point Clouds Based on Semantic Graph and Semantic Scan Context[J]. Laser & Optoelectronics Progress, 2024, 61(18): 1815002

    Download Citation

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

    Category: Machine Vision

    Received: Jan. 2, 2024

    Accepted: Feb. 18, 2024

    Published Online: Sep. 14, 2024

    The Author Email: Yuhui Peng (pengyuhui@fzu.edu.cn)

    DOI:10.3788/LOP240433

    CSTR:32186.14.LOP240433

    Topics