Deep Learning for the Detection and Recognition of Rail Defects in Ultrasound B-Scan Images

Rail defect detection is crucial to rail operations safety. Addressing the problem of high false alarm rates and missed detection rates in rail defect detection, this paper proposes a deep learning method using B-scan image recognition of rail defects with an improved YOLO (you only look once) V3 algorithm. Specifically, the developed model can automatically position a box in B-scan images and recognize EFBWs (electric flash butt welds), normal bolt holes, BHBs (bolt hole breaks), and SSCs (shells, spalling, or corrugation). First, the network structure of the YOLO V3 model is modified to enlarge the receptive field of the model, thus improving the detection accuracy of the model for small-scale objects. Second, B-scan image data are analyzed and standardized. Third, the initial training parameters of the improved YOLO V3 model are adjusted. Finally, the experiments are performed on 453 B-scan images as the test data set. Results show that the B-scan image recognition model based on the improved YOLO V3 algorithm reached high performance in its precision. Additionally, the detection accuracy and efficiency are improved compared with the original model and the final mean average precision can reach 87.41%.

• Record URL:
  https://doi.org/10.1177/03611981211021547
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/03611981
• Supplemental Notes:
  • Zhengxing Chen, Qihang Wang, Kanghua Yang, Ping Wang, and Qing He are now affiliated to School of Civil Engineering, Southwest Jiaotong University, Chengdu, Sichuan, China. © National Academy of Sciences: Transportation Research Board 2021.
• Authors:
  • Chen, Zhengxing
  • Wang, Qihang
  • 0000-0001-7427-8716
  • Yang, Kanghua
  • Yu, Tianle
  • Yao, Jidong
  • Liu, Yong
  • Wang, Ping
  • He, Qing
  • 0000-0003-2596-4984
• Publication Date: 2021-11

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; Photos; References; Tables;
• Pagination: pp 888-901
• Serial:
  • Transportation Research Record: Journal of the Transportation Research Board
  • Volume: 2675
  • Issue Number: 11
  • Publisher: Sage Publications, Incorporated
  • ISSN: 0361-1981
  • EISSN: 2169-4052
  • Serial URL: http://journals.sagepub.com/home/trr

Subject/Index Terms

• TRT Terms: Algorithms; Cluster analysis; Defects; Detection and identification system applications; Image analysis; Machine learning; Maintenance of way; Neural networks; Railroad rails
• Identifier Terms: YOLO
• Subject Areas: Data and Information Technology; Maintenance and Preservation; Railroads;

Filing Info

• Accession Number: 01764801
• Record Type: Publication
• Report/Paper Numbers: TRBAM-21-00304
• Files: TRIS, TRB, ATRI
• Created Date: Feb 16 2021 4:00PM