Optics and Precision Engineering, Volume. 31, Issue 22, 3345(2023)

Adaptive feature matching network for object occlusion

Lin MAO... Hongyang SU* and Dawei YANG |Show fewer author(s)
Author Affiliations
  • School of Electromechanical Engineering, Dalian Minzu University, Dalian116600, China
  • show less
    Figures & Tables(15)
    Overall network framework
    Schematic diagram of tracking drift
    Schematic diagram of feature map splitting
    Multi-response fractional graph
    AFMN block diagram
    Schematic diagram of feature map vector
    Calculation of point-product similarity
    Classification regression network
    Visual comparison
    Visualization results in the object occlusion scenario
    • Table 1. On the OTB-2015 dataset, AFMN compares to other trackers

      View table
      View in Article

      Table 1. On the OTB-2015 dataset, AFMN compares to other trackers

      跟踪器SuccessPrecision跟踪器SuccessPrecision
      AFMN0.7150.928ToMP-50160.701-
      STMTrack140.7190.934MixFormer-1k90.6960.911
      SBT large80.7190.924SiamRPN++50.6960.914
      SAOT170.7140.926KYS180.695-
      SiamAttn190.7120.926Ocean200.6840.899
      UPDT210.7020.919SiamFC++150.683-
    • Table 2. AFMN compares to other trackers on VOT2018 dataset

      View table
      View in Article

      Table 2. AFMN compares to other trackers on VOT2018 dataset

      跟踪器EAOAR
      AFMN0.4520.5930.156
      STMTrack140.4470.5900.159
      D3S70.4890.6400.150
      Ocean200.4890.5920.117
      SiamAttn190.4700.6300.160
      KYS180.4620.6090.143
      SiamBAN220.4520.5970.178
      PrDiMP-50230.4420.6180.165
      DiMP-50240.4400.5970.153
      Siam R-CNN250.4080.6090.220
      SiamFC++150.4260.5870.183
      SiamRPN++50.4140.6000.234
    • Table 3. AFMN compares to other trackers on GOT-10k and LaSOT dataset

      View table
      View in Article

      Table 3. AFMN compares to other trackers on GOT-10k and LaSOT dataset

      跟踪器GOT-10kLaSOT
      AOSR0.5SR0.75AUCPNormP
      AFMN0.6600.7610.5940.6080.6960.636
      STMTrack140.6420.7370.5750.6060.6930.633
      MixFormer-1k90.7320.8320.7020.6790.7730.739
      STARK260.6880.7810.6410.6710.770-
      SBT large80.7040.8080.6470.667-0.717
      TrDiMP230.6710.7770.5830.639-0.614
      AutoMatch270.6520.7660.5430.582-0.599
      SAOT170.6400.759-0.6160.708-
      Siam R-CNN250.6490.7280.5970.6480.722-
      Ocean200.6110.7210.4730.5600.6510.566
      D3S70.5970.6760.462---
      SiamFC++150.5950.6950.4790.5440.6230.547
      SiamRPN++50.5170.6160.3250.4960.5690.491
    • Table 4. 对的影响

      View table
      View in Article

      Table 4. 对的影响

      h数量AOSR0.5SR0.75
      10.6420.7370.575
      20.6560.7550.586
      30.6600.7610.594
      40.6490.7460.577
      50.6420.7400.570
      80.6390.7370.561
    • Table 5. 对的影响

      View table
      View in Article

      Table 5. 对的影响

      T数量AOSR0.5SR0.75
      10.5980.6380.437
      20.6450.7430.573
      30.6600.7610.594
      40.6340.7280.567
      50.6280.7280.548
      60.6380.7380.566
    Tools

    Get Citation

    Copy Citation Text

    Lin MAO, Hongyang SU, Dawei YANG. Adaptive feature matching network for object occlusion[J]. Optics and Precision Engineering, 2023, 31(22): 3345

    Download Citation

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

    Category:

    Received: Apr. 26, 2023

    Accepted: --

    Published Online: Dec. 29, 2023

    The Author Email: SU Hongyang (wxhxhwdn0725@163.com)

    DOI:10.37188/OPE.20233122.3345

    Topics