Journal of Applied Optics, Volume. 46, Issue 2, 343(2025)

Long-time tracking technology for ground targets based on deep learning

Xiaoyan LU1、*, Meng SHEN2, Jie WANG1, Jiaheng LI1, Yizhou YANG1, Xi HE1, Yuju CAO1, and Lan PANG1
Author Affiliations
  • 1Xi'an Institute of Applied Optics, Xi'an 710065, China
  • 2Xi'an Institute of Modern Control Technology, Xi'an 710065, China
  • show less
    Figures & Tables(20)
    Schematic diagram of ECO algorithm using deep features for correlation filtering
    Schematic diagram of feature map interpolation as continuous domain
    Schematic diagram of changes in response graph
    Framework diagram of multiple detection and anti-occlusion tracking algorithm
    Algorithm flow for anti occlusion based on background features
    Schematic diagram of overall mapping of YOLOv5 model
    Structure diagram of FPN+PAN network
    Schematic diagram of channel pruning principle
    Automatic capture when target reappears after occlusion
    Tracking effect of BlurCar1 sequence
    Tracking effect of BlurOwl sequence
    Tracking effect of bird1 sequence
    Tracking effect of DragonBaby sequence
    Test results of BlurCar3 sequence
    Test results of BlurOwl sequence
    Test results of bird sequence
    Test results of DragonBaby sequence
    • Table 1. Test results of dataset

      View table
      View in Article

      Table 1. Test results of dataset

      序列名称图像分辨率/像素最小目标尺寸/像素序列总帧数IoU准确率
      BlurCar1640×48080×807410.50.997
      0.70.942
      0.90.212
      BlurCar2640×48089×815840.51.000
      0.70.995
      0.90.241
      BlurCar3640×48077×663560.51.000
      0.71.000
      0.90.351
      DragonBaby640×480143×1213790.51.000
      0.70.982
      0.90.296
      Car1320×24017×1510190.50.996
      0.70.762
      0.90.047
      Car2320×24046×399120.51.000
      0.70.970
      0.90.125
      bird1320×24052×446580.51.000
      0.70.979
      0.90.305
      Car24320×24027×2430580.50.999
      0.70.950
      0.90.124
      CarDark320×24029×223920.50.995
      0.70.570
      0.90.094
      CarScale640×27244×212510.51.000
      0.70.988
      0.90.359
      BlurOwl320×24087×389440.50.944
      0.70.766
      0.90.010
    • Table 2. Experimental data

      View table
      View in Article

      Table 2. Experimental data

      测试项时间/ms备注
      跟踪模块周期运行时间2
      跟踪模块初始化时间100初始化目标大小:60×40像素
      识别模块时间6模型大小:720×640像素
    • Table 3. Experimental data

      View table
      View in Article

      Table 3. Experimental data

      测试项时间/ms备注
      跟踪模块周期运行时间10
      跟踪模块初始化时间500初始化目标大小:60×40像素
      识别模块时间33模型大小:720×640像素
    Tools

    Get Citation

    Copy Citation Text

    Xiaoyan LU, Meng SHEN, Jie WANG, Jiaheng LI, Yizhou YANG, Xi HE, Yuju CAO, Lan PANG. Long-time tracking technology for ground targets based on deep learning[J]. Journal of Applied Optics, 2025, 46(2): 343

    Download Citation

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

    Category:

    Received: Nov. 20, 2023

    Accepted: --

    Published Online: May. 13, 2025

    The Author Email: Xiaoyan LU (13363906964@189.cn)

    DOI:10.5768/JAO202546.0202007

    Topics