• Laser & Optoelectronics Progress
  • Vol. 58, Issue 12, 1215006 (2021)
Cuijun Zhang1、2 and Yuhe Zhang1、*
Author Affiliations
  • 1College of Information Engineering, Hebei GEO University, Shijiazhuang, Hebei 0 50031, China
  • 2Hebei Center for Ecological and Environmental Geology Research, Hebei GEO University, Shijiazhuang, Hebei 0 50031, China
  • show less
    DOI: 10.3788/LOP202158.1215006 Cite this Article Set citation alerts
    Cuijun Zhang, Yuhe Zhang. Research on SLAM Loop Closure Detection Method Based on HHO Algorithm[J]. Laser & Optoelectronics Progress, 2021, 58(12): 1215006 Copy Citation Text show less
    Contrast charts of feature point detection of improved FAST algorithm. (a) δ=0.1; (b) δ=0.2; (c) δ=0.3; (d) δ=0.4; (e) δ=0.5
    Fig. 1. Contrast charts of feature point detection of improved FAST algorithm. (a) δ=0.1; (b) δ=0.2; (c) δ=0.3; (d) δ=0.4; (e) δ=0.5
    Comparison of feature extraction effect before and after FAST algorithm improvement.(a) Before improvement; (b) after improvement
    Fig. 2. Comparison of feature extraction effect before and after FAST algorithm improvement.(a) Before improvement; (b) after improvement
    Comparison of detection results before and after FAST algorithm improvement when image brightness is reduced by 50%. (a) Before improvement; (b) after improvement
    Fig. 3. Comparison of detection results before and after FAST algorithm improvement when image brightness is reduced by 50%. (a) Before improvement; (b) after improvement
    Comparison of detection results before and after FAST algorithm improvement when image brightness is doubled. (a) Before improvement; (b) after improvement
    Fig. 4. Comparison of detection results before and after FAST algorithm improvement when image brightness is doubled. (a) Before improvement; (b) after improvement
    Feature extraction comparison of the original algorithm. (a) Original image; (b) brightness is reduced by 50%; (c) brightness is doubled
    Fig. 5. Feature extraction comparison of the original algorithm. (a) Original image; (b) brightness is reduced by 50%; (c) brightness is doubled
    Feature extraction comparison of the improved FAST algorithm. (a) Original image; (b) brightness is reduced by 50%; (c) brightness is doubled
    Fig. 6. Feature extraction comparison of the improved FAST algorithm. (a) Original image; (b) brightness is reduced by 50%; (c) brightness is doubled
    Steps of loop closure detection method based on HHO algorithm
    Fig. 7. Steps of loop closure detection method based on HHO algorithm
    P-R curves of three loop closure detection methods. (a) On KITTI dataset; (b) on freiburg2_desk dataset
    Fig. 8. P-R curves of three loop closure detection methods. (a) On KITTI dataset; (b) on freiburg2_desk dataset
    Fitness change curves on freiburg2_desk dataset. (a) Current frame number is 183; (b) current frame number is 2839
    Fig. 9. Fitness change curves on freiburg2_desk dataset. (a) Current frame number is 183; (b) current frame number is 2839
    Fitness change curves on KITTI dataset. (a) Current frame number is 625; (b) current frame number is 3633
    Fig. 10. Fitness change curves on KITTI dataset. (a) Current frame number is 625; (b) current frame number is 3633
    Time comparison of three loop closure detection methods on freiburg2_desk dataset. (a) Current frame number is 183; (b) current frame number is 2839
    Fig. 11. Time comparison of three loop closure detection methods on freiburg2_desk dataset. (a) Current frame number is 183; (b) current frame number is 2839
    Time comparison of three loop closure detection methods on KITTI dataset. (a) Current frame number is 625; (b) current frame number is 3633
    Fig. 12. Time comparison of three loop closure detection methods on KITTI dataset. (a) Current frame number is 625; (b) current frame number is 3633
    Value of r and |E|Strategy
    r≥0.5 and |E|≥0.5Soft besiege
    r≥0.5 and |E|<0.5Hard besiege
    r<0.5 and |E|≥0.5Soft besiege with progressive rapid dives
    r<0.5 and |E|<0.5Hard besiege with progressive rapid dives
    Table 1. Four different location update strategies in the exploitation phase
    Parameterδ
    0.10.20.30.40.5
    Number of feature points419018611037610384
    Table 2. Number of feature points detected by improved FAST algorithm when setting different δ
    Cuijun Zhang, Yuhe Zhang. Research on SLAM Loop Closure Detection Method Based on HHO Algorithm[J]. Laser & Optoelectronics Progress, 2021, 58(12): 1215006
    Download Citation