• Laser & Optoelectronics Progress
  • Vol. 59, Issue 8, 0815006 (2022)
Yao Chen1, Yunwei Zhang1、2、3、*, Jinhui Lei1、3, and li Li1
Author Affiliations
  • 1Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming , Yunnan 650500, China
  • 2Yunnan Key Laboratory of Artificial Intelligence, Kunming University of Science and Technology, Kunming , Yunnan 650500, China
  • 3Yunnan Key Laboratory of Computer Technology Application, Kunming University of Science and Technology, Kunming , Yunnan 650500, China
  • show less
    DOI: 10.3788/LOP202259.0815006 Cite this Article Set citation alerts
    Yao Chen, Yunwei Zhang, Jinhui Lei, li Li. Automatic Extraction Method for Gait Parameters of Quadruped Walking Based on Computer Vision[J]. Laser & Optoelectronics Progress, 2022, 59(8): 0815006 Copy Citation Text show less
    Video acquisition system
    Fig. 1. Video acquisition system
    Images of quadruped walking
    Fig. 2. Images of quadruped walking
    Flow chart of automatic identification of motion feature
    Fig. 3. Flow chart of automatic identification of motion feature
    Structure of DeeplabV3+
    Fig. 4. Structure of DeeplabV3+
    Multiscale feature fusion module
    Fig. 5. Multiscale feature fusion module
    DUpsampling
    Fig. 6. DUpsampling
    Improved DeeplabV3+
    Fig. 7. Improved DeeplabV3+
    Comparison of effects of semantic segmentation.(a) Original image; (b) PSPNet; (c) Segnet; (d) Unet; (e) DeeplabV3+; (f) improved DeeplabV3+Xception
    Fig. 8. Comparison of effects of semantic segmentation.(a) Original image; (b) PSPNet; (c) Segnet; (d) Unet; (e) DeeplabV3+; (f) improved DeeplabV3+Xception
    Flow chart of motion corner detection and matching
    Fig. 9. Flow chart of motion corner detection and matching
    Schematic diagram of corner swing angle
    Fig. 10. Schematic diagram of corner swing angle
    Local maximum values
    Fig. 11. Local maximum values
    Corner detection of quadrupeds. (a) Corners of rhino without overlap; (b) forelimb corners of rhino with overlap; (c) hindlimb corners of buffalo with overlap; (d) forelimb corners of alpaca with overlap
    Fig. 12. Corner detection of quadrupeds. (a) Corners of rhino without overlap; (b) forelimb corners of rhino with overlap; (c) hindlimb corners of buffalo with overlap; (d) forelimb corners of alpaca with overlap
    Curves of corner distance variation
    Fig. 13. Curves of corner distance variation
    Gait cycle of each limb
    Fig. 14. Gait cycle of each limb
    Motion corner detection and matching
    Fig. 15. Motion corner detection and matching
    Rhino’s corner distance variation curves
    Fig. 16. Rhino’s corner distance variation curves
    Rhino’s gait cycle of each limb
    Fig. 17. Rhino’s gait cycle of each limb
    Buffalo’s corner distance variation curves
    Fig. 18. Buffalo’s corner distance variation curves
    Buffalo’s gait cycle of each limb
    Fig. 19. Buffalo’s gait cycle of each limb
    Alpaca’s corner distance variation curves
    Fig. 20. Alpaca’s corner distance variation curves
    Alpaca’s gait cycle of each limb
    Fig. 21. Alpaca’s gait cycle of each limb
    Gait sequences. (a) Gait sequence of buffalo; (b) gait sequence of rhino; (c) gait sequence of alpaca
    Fig. 22. Gait sequences. (a) Gait sequence of buffalo; (b) gait sequence of rhino; (c) gait sequence of alpaca
    Corner swing angles. (a) Corner swing angle of rhino; (b) corner swing angle of buffalo; (c) corner swing angle of alpaca
    Fig. 23. Corner swing angles. (a) Corner swing angle of rhino; (b) corner swing angle of buffalo; (c) corner swing angle of alpaca
    Corner detection for different scales. (a) R=1.06%; (b) R=3.25%; (c) R=6.24%
    Fig. 24. Corner detection for different scales. (a) R=1.06%; (b) R=3.25%; (c) R=6.24%
    TypeDeeplabV3+Improved DeeplabV3+
    Accuracy /%Deviation value /pixelAccuracy /%Deviation value /pixel
    Rhino80.23884.632
    Buffalo82.43685.927
    Alpaca93.728100.019
    Table 1. Comparison of corner detection accuracy
    TypeLimbRhinoBuffaloAlpaca
    T /sf /HzT /sf /HzT /sf /Hz

    Method of this article

    Forelimb 12.000.501.370.731.280.78
    Forelimb 22.330.431.390.721.280.78
    Hindlimb 12.100.481.380.721.280.78
    Hindlimb 21.870.531.390.721.280.78
    Average value2.080.491.380.721.280.78

    Manual calculation

    Forelimb 11.970.511.380.721.260.79
    Forelimb 22.130.471.390.721.270.79
    Hindlimb 12.070.481.360.741.300.77
    Hindlimb 21.930.521.380.721.210.82
    Average value2.070.501.380.731.260.79
    Error /%0.502.000.001.371.561.27
    Table 2. Values of cycle and frequency
    LimbRhinoBuffaloAlpaca
    Method of this articleManual calculationMethod of this articleManual calculationMethod of this articleManual calculation
    Forelimb 11.431.351.151.120.550.57
    Forelimb 21.331.221.051.100.520.50
    Hindlimb 11.401.451.101.150.500.51
    Hindlimb 21.441.381.131.170.520.54
    Average value1.401.361.111.140.520.53
    Average error /%2.852.631.89
    Table 3. Comparison of stride lengths of rhino, buffalo, and alpaca
    Yao Chen, Yunwei Zhang, Jinhui Lei, li Li. Automatic Extraction Method for Gait Parameters of Quadruped Walking Based on Computer Vision[J]. Laser & Optoelectronics Progress, 2022, 59(8): 0815006
    Download Citation