Optics and Precision Engineering, Volume. 32, Issue 5, 740(2024)

Multispectral image fusion method for surface defect detection of IC devices

Yaohua DENG and Zhihai HUANG*
Author Affiliations
  • College of Mechanical and Electrical Engineering, Guangdong University of Technology, Guangzhou510006, China
  • show less
    Figures & Tables(12)
    Theoretical framework of IC device surface defect detection based on multi-spectral image fusion
    Feature point judgment
    Framework of NSST_VP image fusion principle
    PA-PCNN model neuron structure
    Test bench of visible-infrared image pair acquisition
    Comparison of image registration results
    Comparison of infrared and visible image fusion methods for IC devices
    Loss curves of validation set
    • Table 1. Objective evaluation of image registration

      View table
      View in Article

      Table 1. Objective evaluation of image registration

      图 组ORB算法改进ORB算法
      匹配点对数正确匹配点对数P/%RMSE匹配点对数正确匹配点对数P/%RMSE
      第一组341132.414.98534992.51.52
      第二组31722.616.40484185.41.64
      第三组27414.825.59463984.81.85
      平均值31825.818.99494387.81.67
    • Table 2. Objective evaluation of image fusion quality

      View table
      View in Article

      Table 2. Objective evaluation of image fusion quality

      图 像SDSFENAG
      可见光35.53910.0635.0252.672
      红外77.0727.0687.2011.925
      DTCWT融合41.4575.3896.6041.667
      NSCT融合79.38110.4627.3932.921
      CVT融合47.0929.8686.9323.283
      NSST_VP融合49.88214.7897.4464.294
    • Table 3. Dataset distribution

      View table
      View in Article

      Table 3. Dataset distribution

      缺陷类型数 量
      训练集验证集测试集
      划痕483257126
      异物39714286
      凹坑47218392
      崩边28010953
    • Table 4. Defect detection result

      View table
      View in Article

      Table 4. Defect detection result

      检测方法划痕AP崩边AP凹坑AP异物APmAP@0.5
      YOLOv8s-VS0.761 00.466 20.697 10.483 00.601 8
      YOLOv8s-IR0.304 40.585 30.389 60.912 30.548 4
      YOLOv8s-DTCWT0.653 20.651 20.696 50.803 20.701 0
      YOLOv8s-NSCT0.553 60.787 10.404 20.767 20.628 0
      YOLOv8s-CVT0.612 50.696 00.572 30.841 10.680 5
      YOLOv8s-NSST_VP0.822 10.854 50.762 30.887 20.831 5
    Tools

    Get Citation

    Copy Citation Text

    Yaohua DENG, Zhihai HUANG. Multispectral image fusion method for surface defect detection of IC devices[J]. Optics and Precision Engineering, 2024, 32(5): 740

    Download Citation

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

    Category:

    Received: Oct. 5, 2023

    Accepted: --

    Published Online: Apr. 2, 2024

    The Author Email: Zhihai HUANG (2112101013@mail2.gdut.edu.cn)

    DOI:10.37188/OPE.20243205.0740

    Topics