Acta Photonica Sinica, Volume. 52, Issue 12, 1210003(2023)

Adaptive Information Selection for Infrared Object Tracking with Variable Scale Correlation Filter

Mengyu SUN1, Peng WANG2, Junqi XU1、*, Xiaoyan LI2, Hui GAO2, and Ruohai DI2
Author Affiliations
  • 1School of Opto-electronical Engineering,Xi'an Technological University,Xi'an 710021,China
  • 2Electronic Information Engineering,Xi'an Technological University,Xi'an 710021,China
  • show less
    Figures & Tables(12)
    Overall algorithm architecture
    The images in datasets
    The visualization of features
    The evaluation results of LSOTB-TIR dataset
    The evaluation results of PTB-TIR dataset
    The ablation result of location fusion weight
    The ablation result of the number of scales
    The tracking results on four sequences
    • Table 1. The pseudo-code of algorithm

      View table
      View in Article

      Table 1. The pseudo-code of algorithm

      Algorithm 1 The tracking process for one time-step t

      1:Inputs:last frame weights wt-1,ideal response map y,ideal response map for scale estimation ys,current frame feature f,current frame feature for scale estimation fs,last frame feature for scale estimation fst-1,penalty parameter μ

      2:if i = 1 then

      3:wt-1 ← zeros matrix,Γ← zeros matrix

      4:for i=1:Niteration do

      4:wequation 5fwt-1yΓ

      5:w'equation 6wyΓ

      5:update μ,update Γ

      6:end for

      7:else

      8:Swf

      9:Sswsfs

      10:for i=1:Niteration do

      11:wequation 5fwt-1yΓ

      12:w'equation 6wyΓ

      13:update μ,update Γ

      14:end for

      15:wsequation 11ysfs

      16:end if

    • Table 2. The definition of scenarios

      View table
      View in Article

      Table 2. The definition of scenarios

      ScenarioDefinition
      VSThe videos come from a surveillance camera
      HHThe videos are shotted from a hand-held camera
      DSThe videos are captured from a drone-mounted camera
      VMThe videos come from a vehicle-mounted camera
    • Table 3. The attributes of sequences

      View table
      View in Article

      Table 3. The attributes of sequences

      AttributeDefinition
      TCTwo same intensity targets cross each other
      IVThe target intensity is changed during tracking
      DISExisting the intra-class object near the target
      BCThe background has a similar appearance to the target
      DEFThe target is deformable during tracking
      OCCThe target is partly or fully occluded during tracking
      OVThe target partly or fully leaves the image
      SCThe ratio of the target size is out of the range[0.5,2]
      FMThe target moves more than 20 pixels
      MBThe target is blurry due to the target or camera motion
      LRThe target size is lower than 800 pixels
      ARVThe target aspect ratio is out of the range[0.5,2]
    • Table 4. The results of ablation experiments

      View table
      View in Article

      Table 4. The results of ablation experiments

      NameInformation selectionVariable scale estimationFeature extractorPrecisionNorm. precisionSuccess rateFPS
      LADCF0.6860.6290.56934.43
      ACSDCF0.6830.6170.56535.96
      Ours_10.7070.6350.58939.83
      Ours_20.7130.6390.59138.34
      Ours0.7170.6450.59434.85
    Tools

    Get Citation

    Copy Citation Text

    Mengyu SUN, Peng WANG, Junqi XU, Xiaoyan LI, Hui GAO, Ruohai DI. Adaptive Information Selection for Infrared Object Tracking with Variable Scale Correlation Filter[J]. Acta Photonica Sinica, 2023, 52(12): 1210003

    Download Citation

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

    Category:

    Received: Apr. 28, 2023

    Accepted: Jun. 26, 2023

    Published Online: Feb. 19, 2024

    The Author Email: Junqi XU (jqxu2210@163.com)

    DOI:10.3788/gzxb20235212.1210003

    Topics