Laser & Optoelectronics Progress, Volume. 57, Issue 4, 041512(2020)

Scale-Adaptive Correlation Filter Tracking Algorithm Based on FHOG and LBP Features

Xiaoyue Liu*, Yunming Wang, and Weining Ma
Author Affiliations
  • College of Electrical Engineering, North China University of Science and Technology, Tangshan, Hebei 063210, China
  • show less
    Figures & Tables(11)
    Schematic of FHOG feature extraction
    Extraction principle of LBP feature
    Schematic of target tracking and scale estimation for the current frame
    Comparison of accuracy rate
    Comparison of success rate
    Diagrams of partial tracking effects of 6 algorithms
    • Table 1. Information description of the selected image sequence

      View table

      Table 1. Information description of the selected image sequence

      Image sequenceNumber of framesSize /(pixel×pixel)Property description
      Bird1408720×400DEF, FM, OV
      Box1161640×480IV, SV, OCC, MB, IPR, OPR, OV, BC, LR
      Skating2473640×352SV, OCC, DEF, FM, OPR
      Basketball725576×432IV, OCC, DEF, OPR, BC
      Soccer392640×360IV, SV, OCC, MB, FM, IPR, OPR, BC
      CarScale252640×272SV, OCC, FM, IPR, OPR
    • Table 2. Main parameters of the algorithm in this paper

      View table

      Table 2. Main parameters of the algorithm in this paper

      Algorithm parameterValue
      Regularization parameter λ10-4
      Ratio of searching areas1.6
      Learning ratio of apparent model[0.01,0.015]
      Number of directions of FHOG9
      Cell size of FHOG /(pixel×pixel)4×4
      Learning rate η0.01
      Scale factor a1.02
      Scale space N33
    • Table 3. CLE of six algorithms on six sets of test image sequences

      View table

      Table 3. CLE of six algorithms on six sets of test image sequences

      ImagesequenceCLE /pixel
      LCTStapleStruckCSKKCFOur
      Bird16.9254.5998.4986.0319.7784.375
      Box5.9846.7216.1965.6336.7565.376
      Skating26.9693.0953.0652.6065.1892.482
      Basketball9.7657.3867.3527.26912.8897.198
      Soccer4.0334.934.3564.0173.9973.949
      CarScale6.9276.5116.2815.97510.6845.918
    • Table 4. Comparison of DP data of different algorithms

      View table

      Table 4. Comparison of DP data of different algorithms

      ImagesequenceDP /%
      LCTStapleStruckCSKKCFOur
      Bird1100.0034.9036.5035.2095.00100.00
      Box92.8086.4087.2091.6086.40100.00
      Skating2100.0068.9079.1078.90100.00100.00
      Basketball24.6024.0024.0024.00100.00100.00
      Soccer69.4096.7067.9017.9079.3066.60
      CarScale75.8075.2069.4065.1080.6082.10
      Mean77.1064.3560.6851.1790.2291.45
    • Table 5. Comparison of OP data of different algorithms

      View table

      Table 5. Comparison of OP data of different algorithms

      ImagesequenceOP /%
      LCTStapleStruckCSKKCFOur
      Bird1100.0027.6026.8027.6036.40100.00
      Box30.6046.4052.3063.2074.2062.80
      Skating297.7065.9067.9053.3084.0099.80
      Basketball22.3022.3022.2022.30100.0097.70
      Soccer39.0048.2036.7016.3039.3020.70
      CarScale84.5044.8044.6044.8044.4093.70
      Mean62.3542.5341.7537.9263.0579.17
    Tools

    Get Citation

    Copy Citation Text

    Xiaoyue Liu, Yunming Wang, Weining Ma. Scale-Adaptive Correlation Filter Tracking Algorithm Based on FHOG and LBP Features[J]. Laser & Optoelectronics Progress, 2020, 57(4): 041512

    Download Citation

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

    Category: Machine Vision

    Received: Jul. 9, 2019

    Accepted: Aug. 15, 2019

    Published Online: Feb. 20, 2020

    The Author Email: Xiaoyue Liu (807075070@qq.com)

    DOI:10.3788/LOP57.041512

    Topics