• Acta Optica Sinica
  • Vol. 38, Issue 6, 0615001 (2018)
Yi Zhang1、2, Zhiyu Xiang1、2、*, Shuya Chen1、2, and Shuxia Gu1、2
Author Affiliations
  • 1 Zhejiang Provincial Key Laboratory of Information Processing, Communication and Networking, Hangzhou, Zhejiang 310027, China
  • 2 College of Information Science & Electronic Engineering, Zhejiang University, Hangzhou, Zhejiang 310027, China;
  • show less
    DOI: 10.3788/AOS201838.0615001 Cite this Article Set citation alerts
    Yi Zhang, Zhiyu Xiang, Shuya Chen, Shuxia Gu. Optimization on Visual Odometry under Weak Texture Environment[J]. Acta Optica Sinica, 2018, 38(6): 0615001 Copy Citation Text show less
    Framework of system
    Fig. 1. Framework of system
    Flowchart of optimization algorithm
    Fig. 2. Flowchart of optimization algorithm
    Example of stereo matching in KITTI01 sequence. (a) Features in left image; (b) features in right image
    Fig. 3. Example of stereo matching in KITTI01 sequence. (a) Features in left image; (b) features in right image
    Example of image segmentation in KITTI01 sequence
    Fig. 4. Example of image segmentation in KITTI01 sequence
    Feature plane extracted in an image block
    Fig. 5. Feature plane extracted in an image block
    Example of feature re-matching
    Fig. 6. Example of feature re-matching
    Binocular stereoscopic imaging model
    Fig. 7. Binocular stereoscopic imaging model
    Scenes of KITTI 01 sequence
    Fig. 8. Scenes of KITTI 01 sequence
    Comparison of the number of local stereo match points in left and right images (a) before and (b) after optimization
    Fig. 9. Comparison of the number of local stereo match points in left and right images (a) before and (b) after optimization
    Number of stereo matching points in KITTI 01 sequence
    Fig. 10. Number of stereo matching points in KITTI 01 sequence
    Comparison of error of KITTI 01 sequence
    Fig. 11. Comparison of error of KITTI 01 sequence
    Accuracy comparison of different algorithms in KITTI 11 sequence
    Fig. 12. Accuracy comparison of different algorithms in KITTI 11 sequence
    (a) Electric hunting cart; (b) BumbleBee2 binocular camera
    Fig. 13. (a) Electric hunting cart; (b) BumbleBee2 binocular camera
    Part of the experimental scene. (a) Weak texture environment; (b) moving pedestrians and vehicles; (c) shadow, exposure and other light changes
    Fig. 14. Part of the experimental scene. (a) Weak texture environment; (b) moving pedestrians and vehicles; (c) shadow, exposure and other light changes
    Results of actual scene experiment
    Fig. 15. Results of actual scene experiment
    MethodAverage translation error /%Average rotation error /[(°)/m]
    SOFT20.660.0014
    LG_SLAM0.820.0020
    RotRocc+0.830.0026
    FPVO1.100.0023
    ORB_SLAM21.150.0027
    S-PTAM1.190.0025
    Table 1. Precision comparison of different algorithms in KITTI benchmark
    Operationt /s
    Feature extraction and matching0.0390
    Optimized by local feature0.0052
    Feature tracking and pose optimization0.0380
    Total0.0820
    Table 2. System runtime statistics
    Yi Zhang, Zhiyu Xiang, Shuya Chen, Shuxia Gu. Optimization on Visual Odometry under Weak Texture Environment[J]. Acta Optica Sinica, 2018, 38(6): 0615001
    Download Citation