Laser & Optoelectronics Progress, Volume. 60, Issue 24, 2428002(2023)

Loop Closure Detection Algorithm Based on Single Frame-Submap Descriptor Matching

Lianxin Dong*, Kangnian Wang, and Zhanhua Huang
Author Affiliations
  • Key Laboratory of Optoelectronics Information Technology, Ministry of Education, School of Precision Instrument and Opto-Electronics Engineering, Tianjin University, Tianjin 300072, China
  • show less
    Figures & Tables(17)
    SC descriptor diagram
    Field angle schematic diagram of Livox-Mid70 solid-state LiDAR and Velodyne VLP-16 mechanical radar
    Comparison of SC descriptor of the submap and current frame that forms a loopback. (a) SC descriptor of the submap; (b) SC descriptor of current frame
    Comparison of SC descriptor between current frame and historical frame. (a) SC descriptor of submap; (b) binary descriptor of current frame; (c) submap SC descriptor obtained with Fig.4 (b) as the mask; (d) SC descriptor of current frame
    Pre-registration schematic diagram of two frame point clouds forming a loopback
    Registration results obtained by CFB-ICP
    Flow chart of the loopback algorithm
    Location and equipment of campus data collection. (a) Satellite map of the data collection area, where the five-pointed star is the starting point and the ending point of data collection process, and the rectangular frame is the loopback area; (b) handheld experimental platform
    Comparison of the global map under FR-IOSB dataset between the original FAST-LIO2 algorithm and after adding the loopback algorithm. (a) FAST-LIO2; (b) FAST-LIO2 + loopback algorithm
    Comparison of the local map under FR-IOSB dataset between the original FAST-LIO2 algorithm and after adding the loopback algorithm. (a) (d) scene 1; (b) (e) scene 2; (c) (f) scene 3
    Comparison of trajectories of each original algorithm in campus dataset with adding the loopback algorithm
    Three-dimensional trajectories of each algorithm in campus data set. (a) Before and after FAST-LIO2 adding loopback; (b) before and after FASTER-LIO adding loopback
    Comparison of the global map under campus dataset between the original FAST-LIO2 algorithm and after adding the loopback algorithm. (a) FAST-LIO2; (b) FAST-LIO2+loopback algorithm
    Comparison of the local map under campus dataset between the original FAST-LIO2 algorithm and after adding the loopback algorithm. (a) (d) scene 1; (b) (e) scene 2; (c) (f) scene 3
    • Table 1. Comparison between ICP algorithm and CFB-ICP algorithm

      View table

      Table 1. Comparison between ICP algorithm and CFB-ICP algorithm

      Method of point cloud registrationQuantitiy of successMSE when success /m2MSE /m2Time /ms
      ICP algorithm40.07210.724227
      CFB-ICP algorithm200.0530.071126
    • Table 2. Comparison of the starting point and the ending point errors of each algorithm

      View table

      Table 2. Comparison of the starting point and the ending point errors of each algorithm

      Method of front-endError in x /mError in y /mError in z /mTotal error /m
      FAST-LIO20.3181.0651.9472.242
      FAST-LIO2+loopback algorithm0.1350.1450.1890.274
      FASTER-LIO0.2470.7091.3041.505
      FASTER-LIO+loopback algorithm0.1660.2440.0260.296
    • Table 3. Time-consuming analysis of the main steps of the algorithm

      View table

      Table 3. Time-consuming analysis of the main steps of the algorithm

      Getting SC /μsGetting binary SC /μsAligning binary SC /μsSC similarity judgement /μsGraph optimization /msPoint cloud registration /ms
      1202724141135
    Tools

    Get Citation

    Copy Citation Text

    Lianxin Dong, Kangnian Wang, Zhanhua Huang. Loop Closure Detection Algorithm Based on Single Frame-Submap Descriptor Matching[J]. Laser & Optoelectronics Progress, 2023, 60(24): 2428002

    Download Citation

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

    Category: Remote Sensing and Sensors

    Received: Feb. 22, 2023

    Accepted: Apr. 4, 2023

    Published Online: Nov. 27, 2023

    The Author Email: Dong Lianxin (dlx2zaq@163.com)

    DOI:10.3788/LOP230685

    Topics