Fast detection algorithm for surface defects of metal parts based on YOLOv4-mobilenet network

Shihe Guo; Peisi Zhong; Yongpeng Sun; Liang Li; Chao Zhang

doi:10.1117/12.2625428

12 December 2021 Fast detection algorithm for surface defects of metal parts based on YOLOv4-mobilenet network

Shihe Guo, Peisi Zhong, Yongpeng Sun, Liang Li, Chao Zhang

Proceedings Volume 12127, International Conference on Intelligent Equipment and Special Robots (ICIESR 2021); 1212724 (2021) https://doi.org/10.1117/12.2625428
Event: International Conference on Intelligent Equipment and Special Robots (ICIESR 2021), 2021, Qingdao, China

Abstract

To overcome the limitations brought about by manual visual inspection in the process of detecting surface defects of metal parts, and to improve the speed and accuracy of detection, a fast detection algorithm for surface defects of metal parts based on YOLOv4-mobilenet Network is proposed. Based on the principle of the MobileNetV3 network, this thesis designs a feature extraction network with faster detection by combining deep separable convolution, inverted a residual structure with linear bottlenecks, and a lightweight attention model based on the Squeeze-and-Excitation(SE) networks. Finally, a deep learning model was trained for surface defects of metal parts. The experimental results prove that the algorithm is faster and the model data volume is smaller than the traditional SSD and YOLOv4 algorithms, which can cope with fast detection tasks.

Citation Download Citation

Shihe Guo, Peisi Zhong, Yongpeng Sun, Liang Li, and Chao Zhang "Fast detection algorithm for surface defects of metal parts based on YOLOv4-mobilenet network", Proc. SPIE 12127, International Conference on Intelligent Equipment and Special Robots (ICIESR 2021), 1212724 (12 December 2021); https://doi.org/10.1117/12.2625428

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available