Infrared and Laser Engineering, Volume. 54, Issue 6, 20240528(2025)

Infrared and visible image registration of low-voltage electrical equipment based on homography evaluation

Guangpan HUANG1, Linxuan ZHANG1,2, Feng REN3, Qingsong ZENG1, and Rui ZENG3
Author Affiliations
  • 1School of Electrical Engineering, Xinjiang University, Urumqi 830047, China
  • 2National Computer Integrated Manufacturing System (CIMS) Engineering Research Center, Tsinghua University, Beijing 100084, China
  • 3Xinjiang TBEA Automatic Equipment Co. Ltd., Changji 831100, China
  • show less
    Figures & Tables(18)
    Matching points mark the plot
    Fusion diagram of IR and VS. (a) IR fuses to VS; (b) VS fuses to IR
    Two-branch homography estimation model
    Space aggregation structure. (a) Feature decomposition structure; (b) Multilevel gated aggregation
    Channel aggregation structure diagram
    Feature enhancement structure diagram
    Efficient aggregation of linear Transformer attention graphs
    Feature aggregation module
    Key point matching graph
    Matching effect of each single stressor matrix. (a) Label image; (b) Prediction image; (c) Final prediction chart
    Different matching algorithm renderings
    The effect diagram of the proposed image registration algorithm
    RMSE error plots for different matching methods
    • Table 1. Channel aggregation contrast

      View table
      View in Article

      Table 1. Channel aggregation contrast

      MethodParams/MRMSE valCMR valTime/s
      SE0.6918.7820.8820.138
      ECA0.6809.1170.8600.134
      CHG0.8557.7840.9250.140
    • Table 2. Comparison of matching effect of different network backbones

      View table
      View in Article

      Table 2. Comparison of matching effect of different network backbones

      VS BNkIR BNkRMSE valCMR valParams/MTime/s
      ResNetIRNet8.8100.854179.2160.148
      VSNetResNet8.1900.911178.6800.139
      VGGNetIRNet8.5950.873179.1070.150
      VSNetVGGNet8.6140.874178.5700.1427
      VSNetVSNet8.6340.878179.3150.149
      IRNetIRNet8.5110.892179.3870.162
      VSNetIRNet7.7840.925178.8510.140
    • Table 3. Comparison of attention of different Transformer

      View table
      View in Article

      Table 3. Comparison of attention of different Transformer

      MethodParams/MRMSE valCMR valTime/s
      TF1.0508.4810.9010.152
      L-TF1.0508.1410.9210.143
      EL-TF1.0827.7840.9250.140
    • Table 4. Reversible constraint contrast

      View table
      View in Article

      Table 4. Reversible constraint contrast

      MethodParams/MRMSE valCMR valTime/s
      VS-IR174.6518.8910.9020.100
      IR-VS174.6528.5180.8930.110
      IR=VS178.8517.7840.9250.140
    • Table 5. Comparative experiment of different registration algorithms

      View table
      View in Article

      Table 5. Comparative experiment of different registration algorithms

      EstimationORBSURFSIFTSPSGLGLoftrDGOurs
      RMSE val18.15812.18711.40011.67411.10911.0589.4428.4397.784
      CMR val0.3200.4970.5630.4700.6500.6790.7410.8800.925
      RMSE test18.45214.69311.21112.34111.34011.19610.9739.7437.895
      CMR test0.3120.3710.6080.4420.5890.6550.7170.8300.918
      Time/s0.0410.0230.0790.2770.2140.2490.0310.0960.140
    Tools

    Get Citation

    Copy Citation Text

    Guangpan HUANG, Linxuan ZHANG, Feng REN, Qingsong ZENG, Rui ZENG. Infrared and visible image registration of low-voltage electrical equipment based on homography evaluation[J]. Infrared and Laser Engineering, 2025, 54(6): 20240528

    Download Citation

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

    Category: Optical imaging, display and information processing

    Received: Nov. 12, 2024

    Accepted: --

    Published Online: Jul. 1, 2025

    The Author Email:

    DOI:10.3788/IRLA20240528

    Topics