An adaptive feature reconstruction network for the precise segmentation of surface defects on printed circuit boards

Danqing Kang (), Jianhuang Lai (), Junyong Zhu () and Yu Han ()
Additional contact information
Danqing Kang: Sun Yat-Sen University
Jianhuang Lai: Sun Yat-Sen University
Junyong Zhu: Intelligence Eyes Co., Ltd
Yu Han: Sun Yat-Sen University

Journal of Intelligent Manufacturing, 2023, vol. 34, issue 7, No 20, 3197-3214

Abstract: Abstract Segmentation networks based on deep learning are widely used in the field of industrial vision inspection, including for the precise segmentation of surface defects on printed circuit boards (PCBs). However, most previous studies have focused only on the utilization of defect samples with visible defects and underestimated the value of template samples without surface defects. In fact, template samples can provide sufficient prior information to identify defects and are not difficult to obtain in many manufacturing scenarios. Therefore, an adaptive feature reconstruction network (AFRNet) is proposed in this paper to utilize these two types of samples. Specifically, AFRNet consists of two main components: a Siamese encoder with shared parameters for extracting features from the input sample pair, and a symmetrical feature reconstruction module for adaptively fusing these extracted features. Similar image-level and feature-level fusion schemes, as well as spatial misalignment caused by unaligned sample pairs have been carefully studied. Extensive experiments on a real-world PCB surface-defect dataset confirm the effectiveness of the proposed method, demonstrating that it can significantly improve the segmentation performance of multiple baselines, such as DANet, PSPNet and DeepLabv3.

Keywords: Visual inspection; Defect detection; Deep learning; Segmentation network; Computer vision (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10845-022-02008-w Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:joinma:v:34:y:2023:i:7:d:10.1007_s10845-022-02008-w

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/10845

DOI: 10.1007/s10845-022-02008-w

Access Statistics for this article

Journal of Intelligent Manufacturing is currently edited by Andrew Kusiak

More articles in Journal of Intelligent Manufacturing from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().