• Laser & Optoelectronics Progress
  • Vol. 59, Issue 4, 0410011 (2022)
Junsong Zheng, Hao Guo*, Abiao Li, and Jubai An
Author Affiliations
  • School of Information Science and Technology, Dalian Maritime University, Dalian , Liaoning 116026, China
  • show less
    DOI: 10.3788/LOP202259.0410011 Cite this Article Set citation alerts
    Junsong Zheng, Hao Guo, Abiao Li, Jubai An. Real-Time Tracking of Fast Moving Weak Object Based on Siamese Network[J]. Laser & Optoelectronics Progress, 2022, 59(4): 0410011 Copy Citation Text show less
    Fast moving weak object graphs. (a) Dark objects, similar objects, small objects; (b) dark objects; (c) small objects
    Fig. 1. Fast moving weak object graphs. (a) Dark objects, similar objects, small objects; (b) dark objects; (c) small objects
    Convolution characteristic diagram
    Fig. 2. Convolution characteristic diagram
    Feature attention
    Fig. 3. Feature attention
    Temporal and spatial variation of target and background
    Fig. 4. Temporal and spatial variation of target and background
    Frame diagram of proposed algorithm
    Fig. 5. Frame diagram of proposed algorithm
    Tracking results
    Fig. 6. Tracking results
    Success rate and precision on OTB100/50.(a) OTB100 accuracy rate; (b) OTB100 success rate; (c) OTB50 accuracy rate; (d) OTB50 success rate
    Fig. 7. Success rate and precision on OTB100/50.(a) OTB100 accuracy rate; (b) OTB100 success rate; (c) OTB50 accuracy rate; (d) OTB50 success rate
    Tracking results for different scenes
    Fig. 8. Tracking results for different scenes
    SequenceChallengeTotal framesWeak object frame
    Bolt 2BC、DEF293Similar object:1-293
    SkiingIV、SV、OPR、DEF81Small object:1-26;32-35;38-81
    CarDarkBC、IV393Dark object:258-336
    MotorRollingFM、BC、MB、LR164Dark object:14-16;40-46;99-105
    Matrix

    FM、BC、IV、OCC

    I/OPR、IV、SV

    100

    Small object:1-57;similar object:1-56

    Dark object:3-13;19-30;74-86

    Car 1FM、BC、LR、MBIV、SV1020

    Small object:525-1020;similar object:102-300

    Dark object:251-540;614-799;819-897

    Table 1. Tracking sequence

    Algorithms

    Feature

    Update

    Proposed algorithm

    Conv3+Conv4+Conv5

    Yes

    SiamFC14

    Conv5

    No

    KCF4

    HOG

    No

    MEEM18

    CIE Lab

    Yes

    Struck19

    Haar

    Yes

    DLT20

    SDAE

    Yes

    Table 2. Differences between six algorithms

    Algorithms

    OTB100

    OTB50

    Speed /(frame·s-1

    AUC

    PRE

    AUC

    PRE

    Proposed algorithm

    0.583

    0.783

    0.516

    0.710

    88

    SiamFC-MF

    0.581

    0.774

    0.513

    0.696

    112

    SiamFC

    0.578

    0.770

    0.513

    0.692

    180

    CFnet-conv5

    0.586

    0.711

    0.539

    0.67

    43

    MBST

    0.599

    0.783

    0.536

    0.718

    42.9

    Siam-tri

    0.590

    0.781

    0.531

    0.713

    86.3

    SINT++

    0.574

    0.768

    0.624

    0.839

    1

    Table 3. Comparison of class algorithms
    Junsong Zheng, Hao Guo, Abiao Li, Jubai An. Real-Time Tracking of Fast Moving Weak Object Based on Siamese Network[J]. Laser & Optoelectronics Progress, 2022, 59(4): 0410011
    Download Citation