Skip to main navigation Skip to search Skip to main content

Research on Surface Defect Detection Technology of Long-Distance and Long-Span FAST Cable

  • Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Aiming at the problems of large cable span, small number of defect samples and complex background environment in the FAST cable defect detection task, a set of real-time defect detection algorithms based on convolutional neural network is proposed to realize the accurate location and classification of defects. It can achieve a good detection effect for defects with multiple angles and sizes, especially suitable for medium and long distances. The algorithm is verified on the dataset, and its recognition accuracy can reach 91.7%. Equipped on the hardware inference platform, it fully meets the efficiency and accuracy requirements of FAST cable inspection site, can be used in actual cable quality inspection tasks, and can be widely promoted to real-time defect detection of various high-altitude hanging cables.

Original languageEnglish
Title of host publication2023 9th International Conference on Mechatronics and Robotics Engineering, ICMRE 2023
EditorsYongsheng Ma
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages225-228
Number of pages4
ISBN (Electronic)9781665463263
DOIs
StatePublished - 2023
Event9th International Conference on Mechatronics and Robotics Engineering, ICMRE 2023 - Shenzhen, China
Duration: 10 Feb 202312 Feb 2023

Publication series

Name2023 9th International Conference on Mechatronics and Robotics Engineering, ICMRE 2023

Conference

Conference9th International Conference on Mechatronics and Robotics Engineering, ICMRE 2023
Country/TerritoryChina
CityShenzhen
Period10/02/2312/02/23

Keywords

  • FAST
  • cables
  • deep learning
  • defect detection

Fingerprint

Dive into the research topics of 'Research on Surface Defect Detection Technology of Long-Distance and Long-Span FAST Cable'. Together they form a unique fingerprint.

Cite this