Optics and Precision Engineering, Volume. 31, Issue 6, 962(2023)

Design of channel attention network and system for micro target measurement

Yangwei FU1...2, Jin ZHANG1,2,3,*, Zhenxi SUN1,2, Rui ZHANG1,2, Weishi LI1,2,3, and Haojie XIA1,23 |Show fewer author(s)
Author Affiliations
  • 1School of Instrument Science and Opto-electronics Engineering, Hefei University of Technology, Hefei230009, China
  • 2Anhui Province Key Laboratory of Measuring Theory and Precision Instrument, Hefei30009, China
  • 3Engineering Research Center of Safety Critical Industrial Measurement and Control Technology, Ministry of Education, Hefei20009, China
  • show less
    Figures & Tables(18)
    System schematic diagram.
    System device diagram.
    Overall network architecture of proposed MRCA model.
    Attention mechanism module.
    Subpixel convolution module.
    Super-resolution results comparison of different images for scale factor × 2
    Experimental system diagram
    Comparison diagram of USAF1951 resolution plate reconstruction experiment
    Canny edge detection
    Sobel edge detection
    Robert edge detection
    Comparison before and after reconstruction
    Relative error curve of linewidth measurement.
    • Table 1. Average PSNR/SSIM for scale factors ×2 and ×4 on datasets Set5, Set14

      View table
      View in Article

      Table 1. Average PSNR/SSIM for scale factors ×2 and ×4 on datasets Set5, Set14

      算法尺度Set5Set14
      PSNRSSIMPSNRSSIM
      Bicubic

      ×2

      33.660.929 930.240.868 8
      SRCNN36.660.954 232.420.906 3
      VDSR37.530.958 733.030.912 4
      Ours37.800.964 733.620.927 7
      Bicubic

      ×4

      28.420.810 426.000.702 7
      SRCNN30.490.862 827.500.751 3
      VDSR31.350.883 828.010.767 4
      Ours31.560.889 728.210.786 6
    • Table 2. Comparison of objective indexes reconstructed by different algorithms

      View table
      View in Article

      Table 2. Comparison of objective indexes reconstructed by different algorithms

      算法PSNRSSIM参数量
      Bicubic34.950.925 2-
      SRCNN37.550.934 557 K
      VDSR37.750.935 8665 K
      LatticeNet37.900.937 0756 K
      Ours37.860.937 2658 K
    • Table 3. TMSR values of different algorithms and scales

      View table
      View in Article

      Table 3. TMSR values of different algorithms and scales

      算法

      尺度

      LRBicubicSRCNNVDSROurs
      ×21.081.241.120.890.78
      ×31.261.351.211.130.97
      ×41.631.861.661.581.47
    • Table 4. Line spacing pixels

      View table
      View in Article

      Table 4. Line spacing pixels

      组别像元/pixels像素当量/(μm·pixel-1
      原图SR(×2)原图SR(×2)
      水平方向 114.91529.990134.09366.689
      水平方向 214.86529.905134.54466.878
      水平方向 314.80029.755135.13567.216
      垂直方向115.09529.812132.49467.087
      垂直方向215.45830.339129.38365.922
      垂直方向315.16030.122131.92666.397
    • Table 5. Linewidth calculation value and relative error

      View table
      View in Article

      Table 5. Linewidth calculation value and relative error

      组别标准值/μm原图SR(×2)
      水平方向线宽/μm相对误差/%垂直方向线宽/μm相对误差/%水平方向线宽/μm相对误差/%垂直方向线宽/μm相对误差/%
      (-2,2)1 781.801 795.440.771 756.691.141 795.850.791 792.170.58
      (-2,3)1 587.401 605.001.111 557.821.861 606.101.181 588.280.06
      (-2,4)1 414.211 402.770.811 390.781.661 410.510.261 412.960.09
      (-2,5)1 259.921 240.591.531 223.422.301 248.880.881 258.920.08
      (-2,6)1 122.461 096.922.281 082.633.551 092.102.701 099.062.08
      (-1,1)1 000.00983.191.68959.244.08993.550.65994.540.55
      (-1,2)890.90884.450.72867.592.62
      (-1,3)793.70763.483.81780.181.70
      (-1,4)707.11697.561.35669.015.39
      (-1,5)629.96606.203.77585.267.10
      (-1,6)581.23519.710.59513.4712.69
    Tools

    Get Citation

    Copy Citation Text

    Yangwei FU, Jin ZHANG, Zhenxi SUN, Rui ZHANG, Weishi LI, Haojie XIA. Design of channel attention network and system for micro target measurement[J]. Optics and Precision Engineering, 2023, 31(6): 962

    Download Citation

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

    Category: Information Sciences

    Received: Aug. 2, 2022

    Accepted: --

    Published Online: Apr. 4, 2023

    The Author Email: ZHANG Jin (zhangjin@hfut.edu.cn)

    DOI:10.37188/OPE.20233106.0962

    Topics