Small sample PCB defect detection based on meta feature enhancement

Yulin Wang; Xinli Qian; Tao Song; Gang Mou; Xiaoling Xu; Xuehu Liu

doi:10.1117/12.3034120

13 June 2024 Small sample PCB defect detection based on meta feature enhancement

Yulin Wang, Xinli Qian, Tao Song, Gang Mou, Xiaoling Xu, Xuehu Liu

Author Affiliations +

Proceedings Volume 13180, International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2024); 131803U (2024) https://doi.org/10.1117/12.3034120
Event: International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2024), 2024, Guangzhou, China

Abstract

Aiming at PCB surface defect detection tasks under small sample conditions, a meta learning scheme is introduced to fully extract prior knowledge and quickly generalize on new defects. Firstly, combining meta learning with fine-tuning strategies, only fine-tuning the detector head during the meta testing phase to improve classification ambiguity during knowledge transfer; Secondly, to address the issue of confusion between new and base class defects in PCB, a global feature fusion module is designed in support branches to fuse global channel features with original support features to distinguish different defect categories; Finally, introducing a self attention module on the query branch enhances the network's attention to small targets, helping to solve the problem of missed detection of defective targets. The experimental results show that the proposed method exhibits excellent detection performance in 10 shot tasks, achieving 62.4% mAP in the new class.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Yulin Wang, Xinli Qian, Tao Song, Gang Mou, Xiaoling Xu, and Xuehu Liu "Small sample PCB defect detection based on meta feature enhancement", Proc. SPIE 13180, International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2024), 131803U (13 June 2024); https://doi.org/10.1117/12.3034120

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