• Laser & Optoelectronics Progress
  • Vol. 55, Issue 6, 061502 (2018)
Xiaohong Ma*
Author Affiliations
  • Electrical and Electronic Experiment Teaching Center, Shannxi University of Technology, Hanzhong, Shaanxi 723000, China
  • show less
    DOI: 10.3788/LOP55.061502 Cite this Article Set citation alerts
    Xiaohong Ma. Updating Method of Improved Gradient Threshold in Object Tracking[J]. Laser & Optoelectronics Progress, 2018, 55(6): 061502 Copy Citation Text show less
    Comparison of CLE between KCF and the proposed algorithms based on four videos. (a) BlurBody; (b) couple; (c) deer; (d) soccer
    Fig. 1. Comparison of CLE between KCF and the proposed algorithms based on four videos. (a) BlurBody; (b) couple; (c) deer; (d) soccer
    Comparison of APCE and its threshold, and APCE gradient and its threshold (take the couple video as an example). (a) APCE curve of the proposed algorithm; (b) DifAPCE curve of the proposed algorithm; (c) APCE curve of the proposed algorithm (without the judgment of gradient threshold); (d) two frame changes corresponding to the label ring 4 in Fig. 2(b)
    Fig. 2. Comparison of APCE and its threshold, and APCE gradient and its threshold (take the couple video as an example). (a) APCE curve of the proposed algorithm; (b) DifAPCE curve of the proposed algorithm; (c) APCE curve of the proposed algorithm (without the judgment of gradient threshold); (d) two frame changes corresponding to the label ring 4 in Fig. 2(b)
    Comparison of tracking performance with and without gradient threshold judgment. (a) Without gradient threshold; (b) with gradient threshold
    Fig. 3. Comparison of tracking performance with and without gradient threshold judgment. (a) Without gradient threshold; (b) with gradient threshold
    Qualitative tracking comparison of KCF (top) and proposed algorithms (bottom). (a) BlurBody; (b) couple; (c) soccer; (d) deer
    Fig. 4. Qualitative tracking comparison of KCF (top) and proposed algorithms (bottom). (a) BlurBody; (b) couple; (c) soccer; (d) deer
    VideoKCFProposed algorithm
    DP /%Speed /(frame·s-1)DP /%Speed /(frame·s-1)
    Bird16.9176.316.9163.61
    Bird247.555.3649.555.27
    BlurBody58.451.4678.752.76
    Carscale80.6249.7980.6210.13
    Coke83.894.2084.983.85
    Couple25.7171.9763.6161.76
    Deer81.745.184.542.67
    Dragon Baby33.688.3032.191.50
    Dudek87.745.2787.744.59
    Ironman21.7110.6521.4106.71
    Jumping33.9242.9937.7215.77
    Lemming49.553.0749.552.36
    Liquor95.990.9695.285.49
    Matrix17.0152.4618.0133.38
    Motor Rolling4.370.954.968.42
    Soccer79.350.6086.550.25
    Surfer91.0391.0288.6347.65
    Tiger197.543.2397.539.84
    Tiger235.668.0235.967.76
    Vase79.3113.4084.1118.36
    Average value55.5118.2659.4109.53
    Table 1. Comparison of accuracy and speed between KCF and the proposed algorithm (threshold value is 20 pixel)
    TrackerDP /%Speed /(frame·s-1)
    Without gradient threshold57.9166.01
    With gradient threshold63.6161.76
    Table 2. Comparison of tracking precision with and without gradient threshold judgment
    Xiaohong Ma. Updating Method of Improved Gradient Threshold in Object Tracking[J]. Laser & Optoelectronics Progress, 2018, 55(6): 061502
    Download Citation