Laser & Optoelectronics Progress, Volume. 56, Issue 6, 061005(2019)

Nanoparticle Segmentation Based on U-Net Convolutional Neural Network

Fang Zhang1,2, Yue Wu1, Zhitao Xiao1,2、*, Lei Geng1,2, Jun Wu1,2, Yanbei Liu1,2, and Wen Wang1,2
Author Affiliations
  • 1 School of Electronics and Information Engineering, Tianjin Polytechnic University, Tianjin 300387, China
  • 2 Tianjin Key Laboratory of Optoelectronic Detection Technology and System, Tianjin 300387, China
  • show less
    Figures & Tables(9)
    Spherical nanoparticle image taken by TEM and its partially enlarged view. (a) TEM image; (b) partially enlarged view
    Filtering results by partial differential equation. (a) Original nanoparticle image; (b) filtering result
    Network structure with BN
    Comparison chart of loss curves
    Comparison chart 1 of segmentation effect. (a) Level set algorithm; (b) PixelNet; (c) original U-Net; (d) improved network
    Comparison chart 2 of segmentation effect. (a) Level set algorithm; (b) PixelNet; (c) original U-Net; (d) improved network
    Comparison chart 3 of segmentation effect. (a) Level set algorithm; (b) PixelNet; (c) original U-Net; (d) improved network
    • Table 1. Objective indexes of testing results

      View table

      Table 1. Objective indexes of testing results

      MethodNumber ofsamplesMeanRIOUMeanRACC
      Level set method34700.85620.8775
      PixelNet34700.91870.9405
      Original U-Net34700.92220.9472
      Improved U-Net34700.94020.9630
    • Table 2. Comparison of average time consumption

      View table

      Table 2. Comparison of average time consumption

      MethodLevelsetmethodPixelNetOriginalU-NetImprovedU-Net
      Timeconsumption /ms438315276288
    Tools

    Get Citation

    Copy Citation Text

    Fang Zhang, Yue Wu, Zhitao Xiao, Lei Geng, Jun Wu, Yanbei Liu, Wen Wang. Nanoparticle Segmentation Based on U-Net Convolutional Neural Network[J]. Laser & Optoelectronics Progress, 2019, 56(6): 061005

    Download Citation

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

    Category: Image Processing

    Received: Sep. 18, 2018

    Accepted: Oct. 17, 2018

    Published Online: Jul. 30, 2019

    The Author Email: Xiao Zhitao (xiaozhitao@tjpu.edu.cn)

    DOI:10.3788/LOP56.061005

    Topics