Opto-Electronic Engineering, Volume. 52, Issue 3, 240275(2025)

Construction of convolutional neural network model for micro-scale bump on metal pipe fittings

Zihao Liu1,2, Guohao Tao3, Feng Xue4, Yebo Lu5, and Jun Yang2、*
Author Affiliations
  • 1College of Mechanical Engineering, Tianjin University, Tianjin 300072, China
  • 2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314100, China
  • 3School of Computer Science and Technology, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China
  • 4Zhejiang Master Hydraulic Fittings Co., Ltd., Jiaxing, Zhejiang 316002, China
  • 5School of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314100, China
  • show less

    The low detection rate of tiny defects on the surface of metal pipe fittings is a key issue confronting industrial component inspection. In aiming at this problem, an improved YOLOv9-MM model was constructed to improve the accuracy of small target detection. A real-time image acquisition system for precision metal pipe fittings was designed. By using an annular light source combined with a telecentric lens, the surface of pipe fittings can be snapped by the CCD camera and covered at all angles to eliminate the problem of missing areas. The feature map extracted methods of shallow network were introduced, and the upper sampling module of Dysample was combined to realize the dynamic fusion of depth features. By improving the loss function, the precision of small target detection is greatly improved. The results show that the proposed method has an average detection accuracy of 70.2% and a detection speed of 90 f/s. The proposed method shows some feasibility in the actual application.

    Keywords
    Tools

    Get Citation

    Copy Citation Text

    Zihao Liu, Guohao Tao, Feng Xue, Yebo Lu, Jun Yang. Construction of convolutional neural network model for micro-scale bump on metal pipe fittings[J]. Opto-Electronic Engineering, 2025, 52(3): 240275

    Download Citation

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

    Category: Article

    Received: Nov. 26, 2024

    Accepted: Feb. 17, 2025

    Published Online: May. 22, 2025

    The Author Email: Jun Yang (杨俊)

    DOI:10.12086/oee.2025.240275

    Topics