Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 59, Issue 24, 2415006(2022)

Real-Time Optical Fiber End Surface Defects Detection Model Based on Lightweight Improved Network

Minyu Song1, Lirong Chen1、*, Jian'an Liang1, Jinpeng Li1, Zhenzhen Niu1, Zhen Wang1, and Lili Bai2
Author Affiliations
  • 1College of Physics and Electronic Engineering, Shanxi University, Taiyuan 030006, Shanxi, China
  • 2College of Aeronautics and Astronautics, Taiyuan University of Technology, Taiyuan 030006, Shanxi, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (18)
    Equations (11)
    References (24)
    Cited By (3)
    Tools
    Paper Information
    Topics
    Figures & Tables(18)
    Diagram of fiber end surface defects

    Fig. 1. Diagram of fiber end surface defects

    Download full size
    General structure of YOLOv5s

    Fig. 2. General structure of YOLOv5s

    Download full size
    Diagram of YOLOv5s substructure

    Fig. 3. Diagram of YOLOv5s substructure

    Download full size
    Basic unit of shuffleNetV2

    Fig. 4. Basic unit of shuffleNetV2

    Download full size
    ShuffleNetV2 unit down sampled in space

    Fig. 5. ShuffleNetV2 unit down sampled in space

    Download full size
    Structure diagram of convolutional block attention module (CBAM)

    Fig. 6. Structure diagram of convolutional block attention module (CBAM)

    Download full size
    Structure diagram of channel attention module

    Fig. 7. Structure diagram of channel attention module

    Download full size
    Structure diagram of spatial attention module

    Fig. 8. Structure diagram of spatial attention module

    Download full size
    Structure diagram of YOLOv5_CS

    Fig. 9. Structure diagram of YOLOv5_CS

    Download full size
    Comparison of mAP changes during training

    Fig. 10. Comparison of mAP changes during training

    Download full size
    Contrast diagram of training loss function

    Fig. 11. Contrast diagram of training loss function

    Download full size
    P-R graph

    Fig. 12. P-R graph

    Download full size
    Detection results comparison of YOLOv5_CS model and YOLOv5s model. (a), (c) YOLOv5s detection results; (b), (d) YOLOv5_CS ditection results

    Fig. 13. Detection results comparison of YOLOv5_CS model and YOLOv5s model. (a), (c) YOLOv5s detection results; (b), (d) YOLOv5_CS ditection results

    Download full size
    Detection results comparison of YOLOv5_CS model and YOLOv5s model. (a), (c) YOLOv5s detection results; (b), (d) YOLOv5_CS detection results

    Fig. 14. Detection results comparison of YOLOv5_CS model and YOLOv5s model. (a), (c) YOLOv5s detection results; (b), (d) YOLOv5_CS detection results

    Download full size

    • Table 1. Comparison of detection results of five models

      View table

      Table 1. Comparison of detection results of five models

      MethodShuffleNetV2CBAMDelete convolution kernelmAP /%Infertime /ms
      YOLOv5s82.1011.5
      YOLOv5s_A√84.1812.5
      YOLOv5s_B√81.709.1
      YOLOv5s_C√√83.939.9
      YOLOv5_CS√√√83.808.5

    • Table 2. Complexity comparison of five models

      View table

      Table 2. Complexity comparison of five models

      MethodModel of capacity /MNumber of model parameters /MFloating point operations /G
      YOLOv5s14.07.116.3
      YOLOv5s_A16.28.317.4
      YOLOv5s_B7.83.98.5
      YOLOv5s_C7.94.08.6
      YOLOv5_CS2.81.33.8

    • Table 3. Comparison of detection speed of different graphics cards

      View table

      Table 3. Comparison of detection speed of different graphics cards

      MethodTesla T4Tesla K80
      YOLOv5s86.932.0
      YOLOv5s_A80.025.0
      YOLOv5s_B109.040.0
      YOLOv5s_C101.036.2
      YOLOv5_CS118.046.0

    • Table 4. Comparison of average precision of three types of defects

      View table

      Table 4. Comparison of average precision of three types of defects

      MethodDigPitScratch
      YOLOv5s75.179.591.6
      YOLOv5_CS77.781.492.5
    Tools

    Get Citation

    Copy Citation Text

    Minyu Song, Lirong Chen, Jian'an Liang, Jinpeng Li, Zhenzhen Niu, Zhen Wang, Lili Bai. Real-Time Optical Fiber End Surface Defects Detection Model Based on Lightweight Improved Network[J]. Laser & Optoelectronics Progress, 2022, 59(24): 2415006

    Download Citation

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

    Category: Machine Vision

    Received: Sep. 30, 2021

    Accepted: Nov. 1, 2021

    Published Online: Nov. 28, 2022

    The Author Email: Chen Lirong (clr@sxu.edu.cn)

    DOI:10.3788/LOP202259.2415006

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology