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. 58, Issue 4, 0410020(2021)

Wheelset Tread Defect Detection Method Based on Target Detection Network

Li Zhang1, Danping Huang1、*, Shipeng Liao2, Shaodong Yu1,3, Jianqiu Ye1, Xin Wang1, and Na Dong1
Author Affiliations
  • 1School of Mechanical Engineering, Sichuan University of Science & Engineering, Yibin, Sichuan 644000, China;
  • 2Chengdu Institute of Computer Application, Chinese Academy of Sciences, Chengdu, Sichuan 610041, China
  • 3School of Mechanical Engineering, Sichuan University, Chengdu, Sichuan 610065, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (19)
    Equations (0)
    References (22)
    Cited By (2)
    Tools
    Paper Information
    Topics
    Figures & Tables(19)
    Schematic of collection equipment

    Fig. 1. Schematic of collection equipment

    Download full size
    Collected wheelset images

    Fig. 2. Collected wheelset images

    Download full size
    Flow chart of dual network detection algorithm

    Fig. 3. Flow chart of dual network detection algorithm

    Download full size
    Areas with contrasting gray values

    Fig. 4. Areas with contrasting gray values

    Download full size
    Direction-corrected wheelset images

    Fig. 5. Direction-corrected wheelset images

    Download full size
    SSD network structure

    Fig. 6. SSD network structure

    Download full size
    YOLOv3 network structure

    Fig. 7. YOLOv3 network structure

    Download full size
    Distribution chart between original Anchor Boxes and target size

    Fig. 8. Distribution chart between original Anchor Boxes and target size

    Download full size
    M-YOLOv3 network structure

    Fig. 9. M-YOLOv3 network structure

    Download full size
    Distribution chart between new Anchor Boxes and target size

    Fig. 10. Distribution chart between new Anchor Boxes and target size

    Download full size
    P-R curves of three models when extracting tread

    Fig. 11. P-R curves of three models when extracting tread

    Download full size
    Partial identification renderings of the three networks. (a) SSD300; (b) YOLOv3; (c) YOLOv3-tiny

    Fig. 12. Partial identification renderings of the three networks. (a) SSD300; (b) YOLOv3; (c) YOLOv3-tiny

    Download full size
    P-R curves of five models when identifying defects

    Fig. 13. P-R curves of five models when identifying defects

    Download full size
    Partial identification renderings of five kinds of networks. (a) M-YOLOv3; (b) YOLOv3; (c) YOLOv3-tiny; (d) SSD300; (e) SSD512

    Fig. 14. Partial identification renderings of five kinds of networks. (a) M-YOLOv3; (b) YOLOv3; (c) YOLOv3-tiny; (d) SSD300; (e) SSD512

    Download full size
    Missing and misdetection of some tread defects

    Fig. 15. Missing and misdetection of some tread defects

    Download full size
    P-R curve of two detection methods

    Fig. 16. P-R curve of two detection methods

    Download full size

    • Table 1. Tread detection performance

      View table

      Table 1. Tread detection performance

      NetworkAP /%Time /msIOU /%
      SSD30099.831.883.2
      YOLOv396.944.380.1
      YOLOv3-tiny90.613.973.6

    • Table 2. Defect detection performance

      View table

      Table 2. Defect detection performance

      NetworkM-YOLOv3YOLOv3YOLOv3-tinySSD300SSD512
      AP /%89.990.572.219.342.0
      Time /ms58.763.217.524.546.4

    • Table 3. Defect detection performance of two detection methods

      View table

      Table 3. Defect detection performance of two detection methods

      NetworkAP /%Time /ms
      YOLOv373.292.7
      Exper1 & Exper289.792.3
    Tools

    Get Citation

    Copy Citation Text

    Li Zhang, Danping Huang, Shipeng Liao, Shaodong Yu, Jianqiu Ye, Xin Wang, Na Dong. Wheelset Tread Defect Detection Method Based on Target Detection Network[J]. Laser & Optoelectronics Progress, 2021, 58(4): 0410020

    Download Citation

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

    Category: Image Processing

    Received: Jul. 10, 2020

    Accepted: Aug. 12, 2020

    Published Online: Feb. 24, 2021

    The Author Email: Huang Danping (hdpyx2002@163.com)

    DOI:10.3788/LOP202158.0410020

    Topics

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