Article(id=1319955470463386254, tenantId=1146029695717560320, journalId=1278651732997652489, issueId=1319955375944745964, articleNumber=null, orderNo=null, doi=10.12086/oee.2026.250252, pmid=null, cstr=32245.14.oee.2026.250252, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1756051200000, receivedDateStr=2025-08-25, revisedDate=1761235200000, revisedDateStr=2025-10-24, acceptedDate=1761580800000, acceptedDateStr=2025-10-28, onlineDate=1792631507745, onlineDateStr=2026-10-22, pubDate=1771948800000, pubDateStr=2026-02-25, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1792631507745, onlineIssueDateStr=2026-10-22, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1792631507745, creator=13701087609, updateTime=1792631507745, updator=13701087609, issue=Issue{id=1319955375944745964, tenantId=1146029695717560320, journalId=1278651732997652489, year='2026', volume='53', issue='2', pageStart='250117', pageEnd='1', issueExtLink='null', onlineDate='null', pubDate='1771948800000', pubDateStr='2026-02-25', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1792631485210, creator='13701087609', updateTime=1792631485210, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext=null, issueFiles=null, downloadFileDto=null}, startPage=250252, endPage=1, ext={EN=ArticleExt(id=1319955470664712847, articleId=1319955470463386254, tenantId=1146029695717560320, journalId=1278651732997652489, language=EN, title=PCB-YOLO: an enhanced YOLOv8n for lightweight detection of tiny defects in PCB, columnId=1279511634116841602, journalTitle=Opto-Electronic Engineering, columnName=Article, runingTitle=null, highlight=null, articleAbstract=
Objective

In the production process of printed circuit boards (PCBs), common defects such as mouse_bite, open_circuit, short, spur and spurious_copper are frequently encountered. These defects are characterized by their small size, color similarity to the background, and similar features across different defect types, which make detection challenging and prone to false positives and missed detections. Traditional inspection methods, including manual visual checks and conventional automated optical inspection (AOI), suffer from high error rates, limited adaptability, and dependency on manually selected features. Although deep learning approaches, particularly YOLO-series models, have improved detection robustness, they still face three core limitations in PCB defect inspection: insufficient feature extraction for tiny objects, poor balance between model lightweight and accuracy, and inadequate loss function design for small targets. To address the issues of limited samples, low efficiency, large parameter size, and low accuracy in detecting tiny-defects on PCBs, this paper proposes a lightweight detection model named PCB-YOLO based on the YOLOv8n.

Methods

First, the original high-resolution PCB images with diverse background colors are converted to grayscale to minimize color interference. To avoid losing tiny defects during resizing, a sliding-window slicing strategy is applied, generating 640 pixel×480 pixel sub-images with 100 pixels overlaps. Only slices containing defects are retained, effectively increasing the usable dataset while preserving small-object details. Second, considering the significant proportion of small target defects in PCBs, a Conv_SWS (SimAM with slicing) module is designed by integrating the SimAM attention mechanism and merging slicing operations. The input feature map is divided into local patches, each enhanced by SimAM to emphasize small-defect regions, before being recombined. This module replaces standard convolution in downsampling stages, enhancing multi-scale feature extraction without adding parameters. Third, the original C2f module in YOLOv8 is modified by introducing partial convolution (PConv) to comprehensively compute key feature channels, resulting in the design of the FBC2f module. This enhancement improves feature extraction capability while reducing the number of model parameters. Finally, the standard IoU loss is sensitive to minor localization errors and fails when boxes do not overlap. Thus, a combined loss using normalized Wasserstein distance (NWD) and IoU is proposed. NWD models bounding boxes as 2D Gaussian distributions and measures similarity via Wasserstein distance, providing stable gradients even for non-overlapping tiny objects. The final loss is a weighted sum of NWD and IoU, optimized with a balancing factor α=0.5. The model is trained and evaluated on an augmented PCB defect dataset containing 5172 images with 6282 defect instances. Performance is measured using mAP50, mAP50-95, parameters, FLOPs, and frame per second (FPS). Comparative experiments include ablation studies, attention mechanism comparisons, loss function analysis, and benchmarking against state-of-the-art detectors.

Results and Discussions

Experimental results demonstrate the effectiveness of PCB-YOLO: it achieves 95.6% mAP50 on the PCB defect dataset, outperforming YOLOv8n by 2.9%. It also reduces parameters by 23.6% (from 3.006 M to 2.302 M), FLOPs by 22.2% (from 8.1 G to 6.3 G), and model weight by 1.3 M. Moreover, ablation studies are finished to evaluate each proposed component contributes positively: the Conv_SWS module boosts recall by 5% and mAP50 by 2.4%; the FBC2f module reduces parameters by 23.5% with a moderate mAP50 trade-off. the NWD-IoU loss increases mAP50 by 2.3% and mAP50-95 by 2.8%. The full integration of all three modules achieves the best balance of accuracy and efficiency. Comparative evaluations demonstrate that PCB-YOLO outperforms Faster R-CNN, YOLOv5n, YOLOv10n, YOLOv11n, YOLOv12n, and the recently introduced lightweight PIC2f-YOLO in terms of both accuracy and model size. Additionally, it achieves a real-time inference speed of 160 f/s on GPU and 50 f/s on CPU, meeting the throughput demands of industrial applications. Moreover, PCB-YOLO also demonstrates strong generalization on external datasets, attaining mAP50 scores of 98.1% on the industrial aluminum surface defects dataset and 93.6% on the PKU-Market-PCB dataset, thereby verifying its robustness and transferability. Visual comparisons reveal that PCB-YOLO substantially decreases the rates of missed detections and false alarms relative to YOLOv8n, especially in the case of minor defects like shorts and mouse bites. In summary, the results validate that the Conv_SWS module effectively preserves small-object features during downsampling, the FBC2f module maintains representational capacity while reducing computation, and the NWD-IoU loss stabilizes training and improves localization for miniature targets.

Conclusions

This paper proposes PCB-YOLO, a lightweight and accurate detection model designed for tiny PCB defects. The approach integrates grayscale transformation and adaptive slicing during preprocessing to enhance input data quality. It introduces the Conv_SWS module to improve small-object feature retention and designs the parameter-efficient FBC2f module. Additionally, the NWD-IoU loss is adopted to refine the regression of minor defects. Together, these contributions address key challenges in PCB inspection, including limited sample availability, background variation, small defect size, and the accuracy–complexity trade-off. Experimental results show that PCB-YOLO achieves higher detection accuracy with fewer parameters and faster inference than current leading models. It also demonstrates strong generalization across diverse industrial defect datasets. The framework offers a practical, deployable solution for real-time, high-precision PCB quality inspection in manufacturing. Future work may explore further architectural optimizations and extend the method to other micro-defect detection domains.

, authors=Wenbin Deng1, Yixi Guo1, Yuanqiang Lin2, Yebo Lu1, Jun Yang2, authorsList=Wenbin Deng, Yixi Guo, Yuanqiang Lin, Yebo Lu, Jun Yang, authorCompany=null, correspAuthors=Jun Yang, authorNote=null, correspAuthorsNote=
, copyrightStatement=Copyright © 2026 Opto-Electronic Engineering. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1319955485470606074, articleId=1319955470463386254, tenantId=1146029695717560320, journalId=1278651732997652489, language=CN, title=PCB-YOLO:改进YOLOv8n的PCB微小缺陷轻量化检测方法, columnId=1319955120360636489, journalTitle=光电工程, columnName=研究论文, runingTitle=null, highlight=null, articleAbstract=

针对不同底纹背景颜色的印制电路板 (printed circuit board, PCB)微小缺陷检测样本少、效率低、检测算法参数量大和精度低的问题,本文基于YOLOv8n框架提出了一种轻量化检测模型PCB-YOLO。首先,对原始的多种颜色背景的大幅面PCB数据集进行数据预处理,通过灰度变换和切片操作得到适合YOLOv8输入尺寸的PCB数据和降低颜色背景干扰影响,防止PCB的微小缺陷在缩放的过程中丢失;其次,考虑到PCB存在小目标缺陷的占比较大问题,通过结合SimAM注意力机制和合并切片操作,设计Conv_SWS (SimAM with slicing)模块,替代传统的下采样方式,保留小目标的关键信息,从而增强PCB中微小目标缺陷的特征提取并提高检测准确性;然后,对YOLOv8中原有的C2f模块进行改造,引入部分卷积 (PConv)对关键特征通道进行综合计算,设计FBC2f模块,用于增强特征提取能力并减少模型参数量;最后,采用基于Wasserstein距离的度量标准NWD (normalized Wasserstein distance)与标准IoU (intersection over union)结合改进损失函数来加速模型收敛,进一步提高YOLOv8对微小目标的检测能力。实验结果表明,与原始的YOLOv8n相比,在PCB缺陷检测数据集上的PCB-YOLO模型的参数减少了23.6%,mAP50提高了2.9%,达到95.6%。在工业铝片表面缺陷、PKU-Market-PCB缺陷数据集上的泛化性实验也验证了PCB-YOLO模型的有效性。

, authors=邓文斌1, 郭怡希1, 林源强2, 鹿业波1, 杨俊2, *, *, authorsList=邓文斌, 郭怡希, 林源强, 鹿业波, 杨俊, authorCompany=null, correspAuthors=杨俊, authorNote=

作者贡献

邓文斌:数据集的收集、分析,论文写作;郭怡希:程序编写、实验;林源强:数据预处理;鹿业波:论文修改、提供实验条件;杨俊:指导论文选题、论文修改。

邓文斌 (1985-),男,工学博士,嘉兴大学机械工程学院讲师,研究方向为智能装备开发、机器视觉。E-mail:

杨俊 (1978-),男,工学博士,嘉兴大学人工智能学院副教授,研究方向为智能多媒体信息分析与处理、机器学习、计算机视觉。E-mail:

, correspAuthorsNote=
杨俊,
, copyrightStatement=版权所有©《光电工程》编辑部 2026, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=rpxa8HK9S2J/4yDPchi+fA==, magXml=rtGIRiAw0cc5X6zDGXGG+A==, pdfUrl=null, pdf=NEuVUm4Din2yKE5kzd7W3Q==, pdfFileSize=8619609, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=AOJCRnpXN+teA49akcF9PA==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=XxBnY7MzTc6XWQWQlwSWhA==, mapNumber=null, fund=null)}, authors=[Author(id=1319955485890036483, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wbdeng@zjxu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955487542592261, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955485890036483, language=EN, stringName=Wenbin Deng, firstName=Wenbin, middleName=null, lastName=Deng, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955487710364422, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955485890036483, language=CN, stringName=邓文斌, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1嘉兴大学机械工程学院,浙江 嘉兴 314001, bio={"img":"/atL2g6j3PYiOG8HXkpPrQ==","content":"

邓文斌 (1985-),男,工学博士,嘉兴大学机械工程学院讲师,研究方向为智能装备开发、机器视觉。E-mail:

"}, bioImg=/atL2g6j3PYiOG8HXkpPrQ==, bioContent=

邓文斌 (1985-),男,工学博士,嘉兴大学机械工程学院讲师,研究方向为智能装备开发、机器视觉。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485730652924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=1, ext=[AuthorCompanyExt(id=1319955485739041533, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485747430142, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1嘉兴大学机械工程学院,浙江 嘉兴 314001)])]), Author(id=1319955487882330888, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955487999771402, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955487882330888, language=EN, stringName=Yixi Guo, firstName=Yixi, middleName=null, lastName=Guo, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488096240395, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955487882330888, language=CN, stringName=郭怡希, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1嘉兴大学机械工程学院,浙江 嘉兴 314001, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485730652924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=1, ext=[AuthorCompanyExt(id=1319955485739041533, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485747430142, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1嘉兴大学机械工程学院,浙江 嘉兴 314001)])]), Author(id=1319955488201097997, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955488335315727, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488201097997, language=EN, stringName=Yuanqiang Lin, firstName=Yuanqiang, middleName=null, lastName=Lin, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488427590416, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488201097997, language=CN, stringName=林源强, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2嘉兴大学人工智能学院,浙江 嘉兴 314001, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485818733311, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=2, ext=[AuthorCompanyExt(id=1319955485822927616, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485831316225, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2嘉兴大学人工智能学院,浙江 嘉兴 314001)])]), Author(id=1319955488515670802, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955488607945492, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488515670802, language=EN, stringName=Yebo Lu, firstName=Yebo, middleName=null, lastName=Lu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488742163221, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488515670802, language=CN, stringName=鹿业波, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1嘉兴大学机械工程学院,浙江 嘉兴 314001, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485730652924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=1, ext=[AuthorCompanyExt(id=1319955485739041533, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485747430142, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1嘉兴大学机械工程学院,浙江 嘉兴 314001)])]), Author(id=1319955488817660695, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=juneryoung@zjxu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1319955488897352473, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488817660695, language=EN, stringName=Jun Yang, firstName=Jun, middleName=null, lastName=Yang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488981238554, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488817660695, language=CN, stringName=杨俊, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, *, address=2嘉兴大学人工智能学院,浙江 嘉兴 314001, bio={"img":"nLwPCc3xyx/Gjwzg/h+uow==","content":"

杨俊 (1978-),男,工学博士,嘉兴大学人工智能学院副教授,研究方向为智能多媒体信息分析与处理、机器学习、计算机视觉。E-mail:

"}, bioImg=nLwPCc3xyx/Gjwzg/h+uow==, bioContent=

杨俊 (1978-),男,工学博士,嘉兴大学人工智能学院副教授,研究方向为智能多媒体信息分析与处理、机器学习、计算机视觉。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485818733311, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=2, ext=[AuthorCompanyExt(id=1319955485822927616, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485831316225, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2嘉兴大学人工智能学院,浙江 嘉兴 314001)])])], keywords=[Keyword(id=1319955489102873371, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, orderNo=1, keyword=PCB detection), Keyword(id=1319955489195148060, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, orderNo=2, keyword=tiny object detection), Keyword(id=1319955489266451229, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, orderNo=3, keyword=YOLOv8n), Keyword(id=1319955489346143006, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, orderNo=4, keyword=evaluation metrics), Keyword(id=1319955489488749343, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, orderNo=5, keyword=attention mechanism), Keyword(id=1319955489614578464, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, orderNo=1, keyword=PCB检测), Keyword(id=1319955489719436065, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, orderNo=2, keyword=微小目标检测), Keyword(id=1319955489799127842, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, orderNo=3, keyword=YOLOv8n), Keyword(id=1319955489874625315, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, orderNo=4, keyword=度量标准), Keyword(id=1319955489937539876, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, orderNo=5, keyword=注意力机制)], refs=[Reference(id=1319955497571173204, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=吴一全, 赵朗月, 苑玉彬, 等. 基于机器视觉的PCB缺陷检测算法研究现状及展望[J]. 仪器仪表学报, 2022, 43(8): 1−17., articleTitle=null, refAbstract=null), Reference(id=1319955497642476373, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Y Q, Zhao L Y, Yuan Y B, et al. Research status and the prospect of PCB defect detection algorithm based on machine vision[J]. Chin J Sci Instrum, 2022, 43(8): 1−17., articleTitle=null, refAbstract=null), Reference(id=1319955497738945366, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=董静毅, 吕文涛, 包晓敏, 等. 基于机器视觉的PCB板表面缺陷检测方法研究进展[J]. 浙江理工大学学报, 2021, 45(3): 379−389., articleTitle=null, refAbstract=null), Reference(id=1319955497806054231, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Dong J Y, Lü W T, Bao X M, et al. Research progress of the PCB surface defect detection method based on machine vision[J]. J Zhejiang Sci-Tech Univ, 2021, 45(3): 379−389., articleTitle=null, refAbstract=null), Reference(id=1319955497885746008, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Ebayyeh A A R M A, Mousavi A. A review and analysis of automatic optical inspection and quality monitoring methods in electronics industry[J]. IEEE Access, 2020, 8: 183192−183271., articleTitle=null, refAbstract=null), Reference(id=1319955497965437785, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Tsai D M, Huang C K. Defect detection in electronic surfaces using template-based Fourier image reconstruction[J]. IEEE Trans Compon Package Manuf Technol, 2019, 9(1): 163−172., articleTitle=null, refAbstract=null), Reference(id=1319955498120627034, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Hassanin A A I M, Abd El-Samie F E, El Banby G M. A real-time approach for automatic defect detection from PCBs based on SURF features and morphological operations[J]. Multimed Tools Appl, 2019, 78(24): 34437−34457., articleTitle=null, refAbstract=null), Reference(id=1319955498200318811, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Tsai D M, Lin B T. Defect detection of gold-plated surfaces on PCBs using entropy measures[J]. Int J Adv Manuf Technol, 2002, 20(6): 420−428., articleTitle=null, refAbstract=null), Reference(id=1319955498267427676, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Lu Z S, He Q Q, Xiang X G, et al. Defect detection of PCB based on Bayes feature fusion[J]. J Eng, 2018, 2018(16): 1741−1745., articleTitle=null, refAbstract=null), Reference(id=1319955498334536541, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Hu B, Wang J H. Detection of PCB surface defects with improved faster-RCNN and feature pyramid network[J]. IEEE Access, 2020, 8: 108335−108345., articleTitle=null, refAbstract=null), Reference(id=1319955498401645406, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Zeng N Y, Wu P S, Wang Z D, et al. A small-sized object detection oriented multi-scale feature fusion approach with application to defect detection[J]. IEEE Trans Instrum Meas, 2022, 71: 3507014., articleTitle=null, refAbstract=null), Reference(id=1319955498451977055, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=孙志超, 王博, 张晓玲. 基于可变形残差卷积与伸缩式特征金字塔算法的PCB缺陷检测[J]. 电讯技术, 2023, 63(6): 798−805., articleTitle=null, refAbstract=null), Reference(id=1319955498514891616, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Z C, Wang B, Zhang X L. PCB defect detection based on deformable residual convolution and scalable feature pyramid algorithm[J]. Telecommun Eng, 2023, 63(6): 798−805., articleTitle=null, refAbstract=null), Reference(id=1319955498586194785, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Redmon J, Farhadi A. YOLOv3: an incremental improvement[Z]. arXiv: 1804.02767, 2018. https://doi.org/10.48550/arXiv.1804.02767., articleTitle=null, refAbstract=null), Reference(id=1319955498653303650, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Ultralytics. YOLOv5[EB/OL]. (2025-08-25) https://github.com/ultralytics/YOLOv5., articleTitle=null, refAbstract=null), Reference(id=1319955498728801123, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464–7475. https://doi.org/10.1109/CVPR52729.2023.00721., articleTitle=null, refAbstract=null), Reference(id=1319955498804298596, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Ultralytics[EB/OL]. (2025-08-25) GitHub, 2023. https://github.com/ultralytics/ultralytics., articleTitle=null, refAbstract=null), Reference(id=1319955498867213157, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=何国忠, 梁宇. 基于卷积神经网络的PCB缺陷检测[J]. 图学学报, 2022, 43(1): 21−27., articleTitle=null, refAbstract=null), Reference(id=1319955498938516326, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=He G Z, Liang Y. PCB defect detection based on convolutional neural network[J]. J Graphics, 2022, 43(1): 21−27., articleTitle=null, refAbstract=null), Reference(id=1319955499030791015, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Xia K W, Lv Z L, Liu K. et al. Global contextual attention augmented YOLO with ConvMixer prediction heads for PCB surface defect detection[J]. Sci Rep, 2023, 13(1): 9805., articleTitle=null, refAbstract=null), Reference(id=1319955499110482792, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=李振华, 张雷. 改进YOLOv5的轻量级PCB缺陷检测算法[J]. 无线电工程, 2023, 53(6): 1342−1350., articleTitle=null, refAbstract=null), Reference(id=1319955499198563177, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Z H, Zhang L. Lightweight PCB defect detection algorithm of Improved YOLOv5[J]. Radio Eng, 2023, 53(6): 1342−1350., articleTitle=null, refAbstract=null), Reference(id=1319955499261477738, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Bhattacharya A, Cloutier S G. End-to-end deep learning framework for printed circuit board manufacturing defect classification[J]. Sci Rep, 2022, 12(1): 12559., articleTitle=null, refAbstract=null), Reference(id=1319955499328586603, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Lei L, Li H X, Yang H D. Reliable and lightweight adaptive convolution network for PCB surface defect detection[J]. IEEE Trans Instrum Meas, 2024, 73: 2003208., articleTitle=null, refAbstract=null), Reference(id=1319955499391501164, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Gonzalez R C, Wood R E. 数字图像处理[M]. 阮秋琦, 译. 4版. 北京: 电子工业出版社, 2020., articleTitle=null, refAbstract=null), Reference(id=1319955501090194285, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Gonzalez R C, Wood R E. Digital Image Processing[M]. Ruan Q Q, trans. 4th ed. Beijing: Electronics Industry Press, 2020., articleTitle=null, refAbstract=null), Reference(id=1319955501207634798, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Tong K, Wu Y Q, Zhou F. Recent advances in small object detection based on deep learning: a review[J]. Image Vision Comput, 2020, 97: 103910., articleTitle=null, refAbstract=null), Reference(id=1319955501308298095, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Qiu L L, Zhu W G, Li Y G, et al. YOLO-based multi-class target detection for SAR images[C]//Proceedings of 2023 IEEE 7th Information Technology and Mechatronics Engineering Conference (ITOEC), 2023: 1671–1675. https://doi.org/10.1109/ITOEC57671.2023.10292083., articleTitle=null, refAbstract=null), Reference(id=1319955501417350000, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Jooshin H K, Nangir M, Seyedarabi H. Inception-YOLO: computational cost and accuracy improvement of the YOLOv5 model based on employing modified CSP, SPPF, and inception modules[J]. IET Image Process, 2024, 18(8): 1985−1999., articleTitle=null, refAbstract=null), Reference(id=1319955501488653169, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang S F, Chi C, Yao Y Q, et al. Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 9756–9765. https://doi.org/10.1109/CVPR42600.2020.00978., articleTitle=null, refAbstract=null), Reference(id=1319955501564150642, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang L X, Zhang R Y, Li L D, et al. SimAM: a simple, parameter-free attention module for convolutional neural networks[C]//Proceedings of the 38th International Conference on Machine Learning (ICML), 2021: 11863–11874., articleTitle=null, refAbstract=null), Reference(id=1319955501635453811, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen J R, Kao S H, He H, et al. Run, don't walk: chasing higher FLOPS for faster neural networks[C]//Proceedings of 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 12021–12031. https://doi.org/10.1109/CVPR52729.2023.01157., articleTitle=null, refAbstract=null), Reference(id=1319955501723534196, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang H R, Cheng H, Zhang J T. Faster-PGYOLO: an efficient framework for floating debris detection in inland waters[J]. Vis Comput, 2025, 41(7): 5087−5104., articleTitle=null, refAbstract=null), Reference(id=1319955501794837365, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang J W, Xu C, Yang W, et al. A normalized gaussian Wasserstein distance for tiny object detection[Z]. arXiv: 2110.13389, 2021. https://doi.org/10.48550/arXiv.2110.13389., articleTitle=null, refAbstract=null), Reference(id=1319955501861946230, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Tang Y H, Han K, Guo J Y, et al. GhostNetV2: enhance cheap operation with long-range attention[C]//Proceedings of the 36th Conference on Neural Information Processing Systems (NeurIPS), 2022: 724., articleTitle=null, refAbstract=null), Reference(id=1319955501950026615, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen H T, Wang Y H, Guo J Y, et al. VanillaNet: the power of minimalism in deep learning[C]//Proceedings of the 37th International Conference on Neural Information Processing Systems, 2023: 308., articleTitle=null, refAbstract=null), Reference(id=1319955502029718392, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Q L, Wu B G, Zhu P F, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020: 11531–11539. https://doi.org/10.1109/CVPR42600.2020.01155., articleTitle=null, refAbstract=null), Reference(id=1319955502117798777, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Xia Z F, Pan X R, Song S J, et al. Vision transformer with deformable attention[C]//Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022: 4784–4793. https://doi.org/10.1109/CVPR52688.2022.00475., articleTitle=null, refAbstract=null), Reference(id=1319955502197490554, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Y F, Ren W Q, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146−157., articleTitle=null, refAbstract=null), Reference(id=1319955502314931067, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Tong Z J, Chen Y H, Xu Z W, et al. Wise-IoU: bounding box regression loss with dynamic focusing mechanism[Z]. arXiv: 2301.10051, 2023. https://doi.org/10.48550/arXiv.2301.10051., articleTitle=null, refAbstract=null), Reference(id=1319955502478508924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang H, Zhang S J. Shape-IoU: more accurate metric considering bounding box shape and scale[Z]. arXiv: 2312.17663, 2023. https://doi.org/10.48550/arXiv.2312.17663., articleTitle=null, refAbstract=null), Reference(id=1319955502612726653, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Ren S Q, He K M, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Trans Pattern Anal Mach Intell, 2017, 39(6): 1137−1149., articleTitle=null, refAbstract=null), Reference(id=1319955502692418430, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang A, Chen H, Liu L H, et al. YOLOv10: real-time end-to-end object detection[C]//Proceedings of the 38th International Conference on Neural Information Processing Systems, 2024: 3429., articleTitle=null, refAbstract=null), Reference(id=1319955502763721599, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Tian Y J, Ye Q X, Doermann D. YOLOv12: attention-centric real-time object detectors[Z]. arXiv: 2502.12524, 2025. https://doi.org/10.48550/arXiv.2502.12524., articleTitle=null, refAbstract=null), Reference(id=1319955502851801984, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=胡依伦, 杨俊, 许聪源, 等. PIC2f-YOLO: 金属表面缺陷检测轻量化方法[J]. 光电工程, 2025, 52(1): 240250., articleTitle=null, refAbstract=null), Reference(id=1319955502931493761, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Hu Y L, Yang J, Xu C Y, et al. PIC2f-YOLO: a lightweight method for the detection of metal surface defects[J]. Opto-Electron Eng, 2025, 52(1): 240250., articleTitle=null, refAbstract=null), Reference(id=1319955503027962754, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=46, authorNames=null, journalName=null, refType=null, unstructuredReference=铝片表面工业缺陷目标检测[EB/OL]. https://aistudio.baidu.com/datasetdetail/135640., articleTitle=null, refAbstract=null), Reference(id=1319955503111848836, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=47, authorNames=null, journalName=null, refType=null, unstructuredReference=AI Studio. Surface defect detection for aluminum sheets in industrial applications[EB/OL]. (2025-08-25) https://aistudio.baidu.com/datasetdetail/135640. AI Studio., articleTitle=null, refAbstract=null), Reference(id=1319955503166374789, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=48, authorNames=null, journalName=null, refType=null, unstructuredReference=Ding R W, Dai L H, Li G P, et al. TDD-net: a tiny defect detection network for printed circuit boards[J]. CAAI Trans Intell Technol, 2019, 4(2): 110−116., articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1319955485730652924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=1, ext=[AuthorCompanyExt(id=1319955485739041533, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485747430142, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1嘉兴大学机械工程学院,浙江 嘉兴 314001)]), AuthorCompany(id=1319955485818733311, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=2, ext=[AuthorCompanyExt(id=1319955485822927616, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485831316225, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2嘉兴大学人工智能学院,浙江 嘉兴 314001)])], figs=[ArticleFig(id=1319955490059174693, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.1, caption=Defect types, figureFileSmall=/CZKyhkcTEvLFJbA39izSg==, figureFileBig=AOJCRnpXN+teA49akcF9PA==, tableContent=null), ArticleFig(id=1319955490126283558, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图1, caption=缺陷类型, figureFileSmall=/CZKyhkcTEvLFJbA39izSg==, figureFileBig=AOJCRnpXN+teA49akcF9PA==, tableContent=null), ArticleFig(id=1319955490327610151, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.2, caption=PCB images before and after conversion. (a) PCB diagram with green shading; (b) Image after grayscale processing with green shading;(c) PCB diagram with brown shading; (d) Image after grayscale processing with brown shading, figureFileSmall=4+XrAqe4oTEwErQjoGDnPQ==, figureFileBig=R5bd6SEsOTICE3ALJEQQbg==, tableContent=null), ArticleFig(id=1319955490403107624, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图2, caption=转换前后的PCB图。(a)绿色底纹PCB图;(b)绿色底纹灰度化后图像;(c)棕色底纹PCB图;(d)棕色底纹灰度化后图像, figureFileSmall=4+XrAqe4oTEwErQjoGDnPQ==, figureFileBig=R5bd6SEsOTICE3ALJEQQbg==, tableContent=null), ArticleFig(id=1319955492114383657, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.3, caption=Aspect ratio of defects in the dataset, figureFileSmall=/ZYsAikHMfpfmk/afST4eA==, figureFileBig=HxcynWUCAAiZl+0tFr2baQ==, tableContent=null), ArticleFig(id=1319955492332487466, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图3, caption=数据集中缺陷的宽高比, figureFileSmall=/ZYsAikHMfpfmk/afST4eA==, figureFileBig=HxcynWUCAAiZl+0tFr2baQ==, tableContent=null), ArticleFig(id=1319955492445733675, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.4, caption=The distribution of defect types and sizes in the dataset, figureFileSmall=+McDu0XBehALlfpCQ81I/A==, figureFileBig=imdsLvM/41P4O1FT0lF0/A==, tableContent=null), ArticleFig(id=1319955492542202668, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图4, caption=数据集中缺陷类别与尺寸的分布情况, figureFileSmall=+McDu0XBehALlfpCQ81I/A==, figureFileBig=imdsLvM/41P4O1FT0lF0/A==, tableContent=null), ArticleFig(id=1319955492642865965, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.5, caption=Example of minor defects, figureFileSmall=wo+2JX05I/SE7rEUfWJRcA==, figureFileBig=7ODXgffIcJiFRcMJAM6nwQ==, tableContent=null), ArticleFig(id=1319955492751917870, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图5, caption=微小缺陷样例, figureFileSmall=wo+2JX05I/SE7rEUfWJRcA==, figureFileBig=7ODXgffIcJiFRcMJAM6nwQ==, tableContent=null), ArticleFig(id=1319955492860969775, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.6, caption=YOLOv8 model structure, figureFileSmall=orfO2QBBYqD+EfjF0cyc/A==, figureFileBig=sYPagwgnJao6A2BxLgdDvw==, tableContent=null), ArticleFig(id=1319955492961633072, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图6, caption=YOLOv8模型结构, figureFileSmall=orfO2QBBYqD+EfjF0cyc/A==, figureFileBig=sYPagwgnJao6A2BxLgdDvw==, tableContent=null), ArticleFig(id=1319955493045519153, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.7, caption=PCB-YOLO structure, figureFileSmall=jDHaWXL/a4T2qIC6eSGA5A==, figureFileBig=H4q3I48w2VuHZuuR0etJ7Q==, tableContent=null), ArticleFig(id=1319955493129405234, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图7, caption=PCB-YOLO结构, figureFileSmall=jDHaWXL/a4T2qIC6eSGA5A==, figureFileBig=H4q3I48w2VuHZuuR0etJ7Q==, tableContent=null), ArticleFig(id=1319955493242651443, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.8, caption=SimAM attention mechanism, figureFileSmall=46BPUMydRbmkdFvR1Xv4Iw==, figureFileBig=p/7n/IOE2oEKsN9ZUFJzXA==, tableContent=null), ArticleFig(id=1319955493364286260, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图8, caption=SimAM注意力机制, figureFileSmall=46BPUMydRbmkdFvR1Xv4Iw==, figureFileBig=p/7n/IOE2oEKsN9ZUFJzXA==, tableContent=null), ArticleFig(id=1319955493452366645, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.9, caption=Principle of slicing operation, figureFileSmall=GDd2LhtiHGjRMcmiHctv9g==, figureFileBig=St/8YXKQR4I0Nd84NIY2PA==, tableContent=null), ArticleFig(id=1319955493536252726, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图9, caption=切分操作原理, figureFileSmall=GDd2LhtiHGjRMcmiHctv9g==, figureFileBig=St/8YXKQR4I0Nd84NIY2PA==, tableContent=null), ArticleFig(id=1319955493599167287, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.10, caption=Conv_SWS structure, figureFileSmall=Zu4aWIEKZt2hGsUEuYQFUA==, figureFileBig=QYBxBBJPzuZvvVCGIMCVUQ==, tableContent=null), ArticleFig(id=1319955493678859064, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图10, caption=Conv_SWS模块结构, figureFileSmall=Zu4aWIEKZt2hGsUEuYQFUA==, figureFileBig=QYBxBBJPzuZvvVCGIMCVUQ==, tableContent=null), ArticleFig(id=1319955493762745145, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.11, caption=PConv structure, figureFileSmall=9P7aBSK5Q7cR5+4anPZi1A==, figureFileBig=iKnyVCujLvXHjhjdevCz4g==, tableContent=null), ArticleFig(id=1319955493850825530, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图11, caption=PConv结构, figureFileSmall=9P7aBSK5Q7cR5+4anPZi1A==, figureFileBig=iKnyVCujLvXHjhjdevCz4g==, tableContent=null), ArticleFig(id=1319955493926323003, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.12, caption=FasterNet block structure, figureFileSmall=AI/kmNeRZ1AP7ZVQIv9AEg==, figureFileBig=6oX2F3dPpb5fODW7+RsaGQ==, tableContent=null), ArticleFig(id=1319955494006014780, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图12, caption=FasterNet Block结构, figureFileSmall=AI/kmNeRZ1AP7ZVQIv9AEg==, figureFileBig=6oX2F3dPpb5fODW7+RsaGQ==, tableContent=null), ArticleFig(id=1319955494106678077, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.13, caption=FBC2f structure, figureFileSmall=lAdQxGhtO8OKUQBW92+tsw==, figureFileBig=PUlVbifozkznSoI0xl4/rA==, tableContent=null), ArticleFig(id=1319955494224118591, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图13, caption=FBC2f结构, figureFileSmall=lAdQxGhtO8OKUQBW92+tsw==, figureFileBig=PUlVbifozkznSoI0xl4/rA==, tableContent=null), ArticleFig(id=1319955494316393280, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.14, caption=Comparison of algorithm performance before and after dataset augmentation, figureFileSmall=j5W8WURMnIv1IJxe70DyCw==, figureFileBig=5TnDqRAFu71C3xiPZVDdaA==, tableContent=null), ArticleFig(id=1319955494379307841, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图14, caption=数据集增强前后不同算法性能对比, figureFileSmall=j5W8WURMnIv1IJxe70DyCw==, figureFileBig=5TnDqRAFu71C3xiPZVDdaA==, tableContent=null), ArticleFig(id=1319955494484165442, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Fig.15, caption=Detection instance of different methods under the PCB datasets. (a) Original images; (b) YOLOv8n; (c) PCB-YOLO, figureFileSmall=aOCpbI98vglRgBx5/VtHYg==, figureFileBig=t7pGAcvTues0xAsHX/TPNw==, tableContent=null), ArticleFig(id=1319955494555468611, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=图15, caption=不同方法在PCB缺陷数据集下的检测实例。 (a)原始图像; (b) YOLOv8n; (c) PCB-YOLO, figureFileSmall=aOCpbI98vglRgBx5/VtHYg==, figureFileBig=t7pGAcvTues0xAsHX/TPNw==, tableContent=null), ArticleFig(id=1319955494639354692, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.1, caption=

Light weight comparative experimental results

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%mAP50-95/%Param/MFLOPs/GWeight/M
YOLOv8n (baseline)92.759.63.0068.15.9
Baseline + GhostNet90.350.52.7487.55.4
Baseline + VanillaNet89.146.71.8295.63.7
Baseline + FBC2f90.548.32.3016.34.6
), ArticleFig(id=1319955494702269253, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表1, caption=

轻量化对比实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%mAP50-95/%Param/MFLOPs/GWeight/M
YOLOv8n (baseline)92.759.63.0068.15.9
Baseline + GhostNet90.350.52.7487.55.4
Baseline + VanillaNet89.146.71.8295.63.7
Baseline + FBC2f90.548.32.3016.34.6
), ArticleFig(id=1319955494769378118, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.2, caption=

Comparison results of attention mechanisms

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%mAP50-95/%Param/MFLOPs/GWeight/M
Baseline + ECA93.261.62.5977.95.1
Baseline + DAT93.561.23.7467.96.9
Baseline + SimAM93.860.32.5977.95.1
Baseline + SWS95.162.13.0068.15.9
), ArticleFig(id=1319955494840681287, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表2, caption=

不同注意力机制对比结果

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%mAP50-95/%Param/MFLOPs/GWeight/M
Baseline + ECA93.261.62.5977.95.1
Baseline + DAT93.561.23.7467.96.9
Baseline + SimAM93.860.32.5977.95.1
Baseline + SWS95.162.13.0068.15.9
), ArticleFig(id=1319955494916178760, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.3, caption=

Comparison results of loss function

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelP/%R/%mAP50/%mAP50-95/%Param/M
CIoU92.487.892.259.32.302
EIoU91.585.690.855.62.302
Wise-IoU95.888.593.958.72.302
Shape-IoU95.490.894.560.52.302
NWD-IoU96.391.695.662.32.302
), ArticleFig(id=1319955496640037705, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表3, caption=

不同损失函数对比结果

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelP/%R/%mAP50/%mAP50-95/%Param/M
CIoU92.487.892.259.32.302
EIoU91.585.690.855.62.302
Wise-IoU95.888.593.958.72.302
Shape-IoU95.490.894.560.52.302
NWD-IoU96.391.695.662.32.302
), ArticleFig(id=1319955496757478218, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.4, caption=

Comparison results of α value

, figureFileSmall=null, figureFileBig=null, tableContent=
αP/%R/%mAP50/%mAP50-95/%
0.194.288.993.155.8
0.395.590.294.558.6
0.596.391.695.662.3
0.795.890.794.858.9
0.994.589.693.555.4
), ArticleFig(id=1319955496870724427, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表4, caption=

不同α值对比结果

, figureFileSmall=null, figureFileBig=null, tableContent=
αP/%R/%mAP50/%mAP50-95/%
0.194.288.993.155.8
0.395.590.294.558.6
0.596.391.695.662.3
0.795.890.794.858.9
0.994.589.693.555.4
), ArticleFig(id=1319955496942027596, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.5, caption=

Ablation results

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelP/%R/%mAP50/%mAP50-95/%Param/MFLOPs/GWeight/M
Baseline93.487.892.759.63.0068.15.9
+M189.389.590.548.32.3016.34.6
+M292.892.895.162.13.0068.15.9
+M395.190.495.062.43.0068.15.9
+M1M292.487.892.251.12.3026.34.6
+M1M2M396.391.695.662.32.3026.34.6
), ArticleFig(id=1319955497025913677, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表5, caption=

消融实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelP/%R/%mAP50/%mAP50-95/%Param/MFLOPs/GWeight/M
Baseline93.487.892.759.63.0068.15.9
+M189.389.590.548.32.3016.34.6
+M292.892.895.162.13.0068.15.9
+M395.190.495.062.43.0068.15.9
+M1M292.487.892.251.12.3026.34.6
+M1M2M396.391.695.662.32.3026.34.6
), ArticleFig(id=1319955497093022542, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.6, caption=

Comparison results of different models

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelP/%R/%mAP50/%mAP50-95/%Param/MFLOPs/GWeight/MFPS (CPU)FPS (GPU)
Faster-RCNN82.683.482.344.341.4093.6317.6823
YOLOv5n92.286.392.251.12.507.15.055227
YOLOv10n93.790.193.159.92.718.45.558225
YOLOv11n88.683.987.746.82.586.35.256216
YOLOv12n89.183.488.448.02.566.35.344177
PIC2f-YOLO92.087.492.457.02.6010.65.236152
YOLOv8n93.487.892.759.63.018.15.955229
PCB-YOLO96.391.695.662.32.306.34.650160
), ArticleFig(id=1319955497181102927, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表6, caption=

不同模型对比实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelP/%R/%mAP50/%mAP50-95/%Param/MFLOPs/GWeight/MFPS (CPU)FPS (GPU)
Faster-RCNN82.683.482.344.341.4093.6317.6823
YOLOv5n92.286.392.251.12.507.15.055227
YOLOv10n93.790.193.159.92.718.45.558225
YOLOv11n88.683.987.746.82.586.35.256216
YOLOv12n89.183.488.448.02.566.35.344177
PIC2f-YOLO92.087.492.457.02.6010.65.236152
YOLOv8n93.487.892.759.63.018.15.955229
PCB-YOLO96.391.695.662.32.306.34.650160
), ArticleFig(id=1319955497264989008, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.7, caption=

Generalization experiment results on the aluminum sheet surface industrial defect dataset

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%Param/MWeight/M
YOLOv8n97.33.016.1
PIC2f-YOLO95.02.605.2
PCB-YOLO98.12.304.6
), ArticleFig(id=1319955497327903569, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表7, caption=

铝片表面工业缺陷数据集泛化实验

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%Param/MWeight/M
YOLOv8n97.33.016.1
PIC2f-YOLO95.02.605.2
PCB-YOLO98.12.304.6
), ArticleFig(id=1319955497395012434, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=EN, label=Tab.8, caption=

Generalization experiment results on the PKU-Market-PCB dataset

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%Param/MWeight/M
YOLOv8n90.93.014.6
PIC2f-YOLO92.42.605.2
PCB-YOLO93.62.304.6
), ArticleFig(id=1319955497457926995, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, language=CN, label=表8, caption=

PKU-Market-PCB数据集泛化实验

, figureFileSmall=null, figureFileBig=null, tableContent=
ModelmAP50/%Param/MWeight/M
YOLOv8n90.93.014.6
PIC2f-YOLO92.42.605.2
PCB-YOLO93.62.304.6
)], attaches=null, journal=Journal(id=1278641367198941188, delFlag=0, nameCn=光电工程, nameEn=Opto-Electronic Engineering, nameHistory1=null, nameHistory2=null, issn=1003-501X, eissn=2097-4019, cn=51-1346/O4, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=4Vimkd+qXLWNxtdpr9mFNw==, journalPrice=null, startedYear=null, abbrevIsoEn=Opto-Electronic Engineering, journalRemark=null, publicationField=null, createdTime=1782781457950, updatedTime=1784021784694, createdBy=18614031015, updatedBy=13041195026, firstLetterCn=G, firstLetterEn=G, subjectCode=Engineering, subjectName=null, subjectCodeEn=Engineering, subjectNameEn=null, picCn=4Vimkd+qXLWNxtdpr9mFNw==, picEn=vAq9s20WLs1ODDfbWq+Gjg==, jcr=null, cjcr=null, exts=[JournalExt(id=1283843675721536154, language=CN, name=光电工程, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1784021784954, updatedTime=1784021784954, createdBy=13041195026, updatedBy=13041195026, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.manuscripts.com.cn/gdgc, submissionEditorUrl=http://www.manuscripts.com.cn/gdgc, submissionReviewUrl=http://www.manuscripts.com.cn/gdgc, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1283843675771867803, language=EN, name=Opto-Electronic Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1784021784966, updatedTime=1784021784966, createdBy=13041195026, updatedBy=13041195026, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.manuscripts.com.cn/gdgc, submissionEditorUrl=http://www.manuscripts.com.cn/gdgc, submissionReviewUrl=http://www.manuscripts.com.cn/gdgc, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1278651732997652489, websiteList=[Website(id=1278723867418018151, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1278651732997652489, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/oee/CN, language=CN, createTime=1782801127533, createBy=18614031015, updateTime=1782804494442, updateBy=18614031015, name=光电工程-中文, tplId=1146099689490845704, title=光电工程, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1278738091150128034, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=articleTextType, value=kx, createTime=1782804518735, updateTime=1782804518735, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091120767903, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=banner, value=null, createTime=1782804518728, updateTime=1782804518728, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091171099557, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=grayFlag, value=0, createTime=1782804518740, updateTime=1782804518740, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091108184990, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=logo, value=https://castjournals.cast.org.cn/joweb/oee/CN/file/pic?fileId=A1C6uwqtMazluiWkEpR0Mg==, createTime=1782804518725, updateTime=1782804518725, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091179488167, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=minRunFlag, value=0, createTime=1782804518742, updateTime=1782804518742, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091141739425, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/oee/CN/file/pic, createTime=1782804518733, updateTime=1782804518733, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091175293862, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=silenceFlag, value=0, createTime=1782804518741, updateTime=1782804518741, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091129156512, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1782804518730, updateTime=1782804518730, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091154322339, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=themeColor, value=null, createTime=1782804518736, updateTime=1782804518736, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738091162710948, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867418018151, code=themeStyle, value=null, createTime=1782804518738, updateTime=1782804518738, creator=18614031015, updator=18614031015)]), Website(id=1278723867522875769, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1278651732997652489, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/oee/EN, language=EN, createTime=1782801127558, createBy=18614031015, updateTime=1782804490442, updateBy=18614031015, name=光电工程-英文, tplId=1146101810881728533, title=Opto-Electronic Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1278738063660659607, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=articleTextType, value=kx, createTime=1782804512181, updateTime=1782804512181, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063635493780, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=banner, value=null, createTime=1782804512175, updateTime=1782804512175, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063924900762, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=grayFlag, value=0, createTime=1782804512244, updateTime=1782804512244, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063606133651, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=logo, value=https://castjournals.cast.org.cn/joweb/oee/EN/file/pic?fileId=A1C6uwqtMazluiWkEpR0Mg==, createTime=1782804512168, updateTime=1782804512168, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063937483676, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=minRunFlag, value=0, createTime=1782804512247, updateTime=1782804512247, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063652270998, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/oee/EN/file/pic, createTime=1782804512179, updateTime=1782804512179, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063933289371, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=silenceFlag, value=0, createTime=1782804512246, updateTime=1782804512246, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063639688085, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1782804512176, updateTime=1782804512176, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063664853912, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=themeColor, value=null, createTime=1782804512182, updateTime=1782804512182, creator=18614031015, updator=18614031015), WebsiteProps(id=1278738063916512153, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1278723867522875769, code=themeStyle, value=null, createTime=1782804512242, updateTime=1782804512242, creator=18614031015, updator=18614031015)])], journalTitle=光电工程, weixinUrl=null, journalUrl=https://www.oejournal.org/oee, iacademicId=null, status=1, seqNo=null, journalTitleEn=Opto-Electronic Engineering, journalPhotoCn=4Vimkd+qXLWNxtdpr9mFNw==, journalPhotoEn=vAq9s20WLs1ODDfbWq+Gjg==, journalFirstLetter=G, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/oee/CN/10.12086/oee.2026.250252, detailUrlEn=https://castjournals.cast.org.cn/joweb/oee/EN/10.12086/oee.2026.250252, pdfUrlCn=https://castjournals.cast.org.cn/joweb/oee/CN/PDF/10.12086/oee.2026.250252, pdfUrlEn=https://castjournals.cast.org.cn/joweb/oee/EN/PDF/10.12086/oee.2026.250252, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1771948800000, fullTextJson=null, articleText=null, reference=null)
收藏切换
PCB-YOLO:改进YOLOv8n的PCB微小缺陷轻量化检测方法
收藏切换
PDF下载
邓文斌 1 , 郭怡希 1 , 林源强 2 , 鹿业波 1 , 杨俊 2, *, *
光电工程 | 研究论文 2026,53(2): 250252-1
收起
收藏切换
光电工程 |研究论文 2026 , 53 (2) : 250252 -1
PCB-YOLO:改进YOLOv8n的PCB微小缺陷轻量化检测方法
全屏
[Author(id=1319955485890036483, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wbdeng@zjxu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955487542592261, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955485890036483, language=EN, stringName=Wenbin Deng, firstName=Wenbin, middleName=null, lastName=Deng, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955487710364422, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955485890036483, language=CN, stringName=邓文斌, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1嘉兴大学机械工程学院,浙江 嘉兴 314001, bio={"img":"/atL2g6j3PYiOG8HXkpPrQ==","content":"

邓文斌 (1985-),男,工学博士,嘉兴大学机械工程学院讲师,研究方向为智能装备开发、机器视觉。E-mail:

"}, bioImg=/atL2g6j3PYiOG8HXkpPrQ==, bioContent=

邓文斌 (1985-),男,工学博士,嘉兴大学机械工程学院讲师,研究方向为智能装备开发、机器视觉。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485730652924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=1, ext=[AuthorCompanyExt(id=1319955485739041533, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485747430142, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1嘉兴大学机械工程学院,浙江 嘉兴 314001)])]), Author(id=1319955487882330888, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955487999771402, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955487882330888, language=EN, stringName=Yixi Guo, firstName=Yixi, middleName=null, lastName=Guo, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488096240395, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955487882330888, language=CN, stringName=郭怡希, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1嘉兴大学机械工程学院,浙江 嘉兴 314001, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485730652924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=1, ext=[AuthorCompanyExt(id=1319955485739041533, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485747430142, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1嘉兴大学机械工程学院,浙江 嘉兴 314001)])]), Author(id=1319955488201097997, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955488335315727, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488201097997, language=EN, stringName=Yuanqiang Lin, firstName=Yuanqiang, middleName=null, lastName=Lin, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488427590416, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488201097997, language=CN, stringName=林源强, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2嘉兴大学人工智能学院,浙江 嘉兴 314001, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485818733311, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=2, ext=[AuthorCompanyExt(id=1319955485822927616, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485831316225, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2嘉兴大学人工智能学院,浙江 嘉兴 314001)])]), Author(id=1319955488515670802, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1319955488607945492, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488515670802, language=EN, stringName=Yebo Lu, firstName=Yebo, middleName=null, lastName=Lu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488742163221, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488515670802, language=CN, stringName=鹿业波, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1嘉兴大学机械工程学院,浙江 嘉兴 314001, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485730652924, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=1, ext=[AuthorCompanyExt(id=1319955485739041533, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485747430142, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485730652924, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1嘉兴大学机械工程学院,浙江 嘉兴 314001)])]), Author(id=1319955488817660695, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=juneryoung@zjxu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1319955488897352473, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488817660695, language=EN, stringName=Jun Yang, firstName=Jun, middleName=null, lastName=Yang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1319955488981238554, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, authorId=1319955488817660695, language=CN, stringName=杨俊, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, *, address=2嘉兴大学人工智能学院,浙江 嘉兴 314001, bio={"img":"nLwPCc3xyx/Gjwzg/h+uow==","content":"

杨俊 (1978-),男,工学博士,嘉兴大学人工智能学院副教授,研究方向为智能多媒体信息分析与处理、机器学习、计算机视觉。E-mail:

"}, bioImg=nLwPCc3xyx/Gjwzg/h+uow==, bioContent=

杨俊 (1978-),男,工学博士,嘉兴大学人工智能学院副教授,研究方向为智能多媒体信息分析与处理、机器学习、计算机视觉。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1319955485818733311, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, xref=2, ext=[AuthorCompanyExt(id=1319955485822927616, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China), AuthorCompanyExt(id=1319955485831316225, tenantId=1146029695717560320, journalId=1278651732997652489, articleId=1319955470463386254, companyId=1319955485818733311, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2嘉兴大学人工智能学院,浙江 嘉兴 314001)])])]
邓文斌1 , 郭怡希1, 林源强2, 鹿业波1, 杨俊2, *, *
作者信息
  • 1嘉兴大学机械工程学院,浙江 嘉兴 314001
  • 2嘉兴大学人工智能学院,浙江 嘉兴 314001
通讯作者:
作者简介:

作者贡献

邓文斌:数据集的收集、分析,论文写作;郭怡希:程序编写、实验;林源强:数据预处理;鹿业波:论文修改、提供实验条件;杨俊:指导论文选题、论文修改。

邓文斌 (1985-),男,工学博士,嘉兴大学机械工程学院讲师,研究方向为智能装备开发、机器视觉。E-mail:

杨俊 (1978-),男,工学博士,嘉兴大学人工智能学院副教授,研究方向为智能多媒体信息分析与处理、机器学习、计算机视觉。E-mail:

PCB-YOLO: an enhanced YOLOv8n for lightweight detection of tiny defects in PCB
Wenbin Deng1 , Yixi Guo1, Yuanqiang Lin2, Yebo Lu1, Jun Yang2
Affiliations
  • 1College of Mechanical Engineering, Jiaxing University, Jiaxing, Zhejiang 314001, China
  • 2College of Artificial Intelligence, Jiaxing University, Jiaxing, Zhejiang 314001, China
  • Corresponding Author:
出版时间: 2026-02-25 doi: 10.12086/oee.2026.250252
文章导航
收藏切换

针对不同底纹背景颜色的印制电路板 (printed circuit board, PCB)微小缺陷检测样本少、效率低、检测算法参数量大和精度低的问题,本文基于YOLOv8n框架提出了一种轻量化检测模型PCB-YOLO。首先,对原始的多种颜色背景的大幅面PCB数据集进行数据预处理,通过灰度变换和切片操作得到适合YOLOv8输入尺寸的PCB数据和降低颜色背景干扰影响,防止PCB的微小缺陷在缩放的过程中丢失;其次,考虑到PCB存在小目标缺陷的占比较大问题,通过结合SimAM注意力机制和合并切片操作,设计Conv_SWS (SimAM with slicing)模块,替代传统的下采样方式,保留小目标的关键信息,从而增强PCB中微小目标缺陷的特征提取并提高检测准确性;然后,对YOLOv8中原有的C2f模块进行改造,引入部分卷积 (PConv)对关键特征通道进行综合计算,设计FBC2f模块,用于增强特征提取能力并减少模型参数量;最后,采用基于Wasserstein距离的度量标准NWD (normalized Wasserstein distance)与标准IoU (intersection over union)结合改进损失函数来加速模型收敛,进一步提高YOLOv8对微小目标的检测能力。实验结果表明,与原始的YOLOv8n相比,在PCB缺陷检测数据集上的PCB-YOLO模型的参数减少了23.6%,mAP50提高了2.9%,达到95.6%。在工业铝片表面缺陷、PKU-Market-PCB缺陷数据集上的泛化性实验也验证了PCB-YOLO模型的有效性。

PCB检测  /  微小目标检测  /  YOLOv8n  /  度量标准  /  注意力机制
Objective

In the production process of printed circuit boards (PCBs), common defects such as mouse_bite, open_circuit, short, spur and spurious_copper are frequently encountered. These defects are characterized by their small size, color similarity to the background, and similar features across different defect types, which make detection challenging and prone to false positives and missed detections. Traditional inspection methods, including manual visual checks and conventional automated optical inspection (AOI), suffer from high error rates, limited adaptability, and dependency on manually selected features. Although deep learning approaches, particularly YOLO-series models, have improved detection robustness, they still face three core limitations in PCB defect inspection: insufficient feature extraction for tiny objects, poor balance between model lightweight and accuracy, and inadequate loss function design for small targets. To address the issues of limited samples, low efficiency, large parameter size, and low accuracy in detecting tiny-defects on PCBs, this paper proposes a lightweight detection model named PCB-YOLO based on the YOLOv8n.

Methods

First, the original high-resolution PCB images with diverse background colors are converted to grayscale to minimize color interference. To avoid losing tiny defects during resizing, a sliding-window slicing strategy is applied, generating 640 pixel×480 pixel sub-images with 100 pixels overlaps. Only slices containing defects are retained, effectively increasing the usable dataset while preserving small-object details. Second, considering the significant proportion of small target defects in PCBs, a Conv_SWS (SimAM with slicing) module is designed by integrating the SimAM attention mechanism and merging slicing operations. The input feature map is divided into local patches, each enhanced by SimAM to emphasize small-defect regions, before being recombined. This module replaces standard convolution in downsampling stages, enhancing multi-scale feature extraction without adding parameters. Third, the original C2f module in YOLOv8 is modified by introducing partial convolution (PConv) to comprehensively compute key feature channels, resulting in the design of the FBC2f module. This enhancement improves feature extraction capability while reducing the number of model parameters. Finally, the standard IoU loss is sensitive to minor localization errors and fails when boxes do not overlap. Thus, a combined loss using normalized Wasserstein distance (NWD) and IoU is proposed. NWD models bounding boxes as 2D Gaussian distributions and measures similarity via Wasserstein distance, providing stable gradients even for non-overlapping tiny objects. The final loss is a weighted sum of NWD and IoU, optimized with a balancing factor α=0.5. The model is trained and evaluated on an augmented PCB defect dataset containing 5172 images with 6282 defect instances. Performance is measured using mAP50, mAP50-95, parameters, FLOPs, and frame per second (FPS). Comparative experiments include ablation studies, attention mechanism comparisons, loss function analysis, and benchmarking against state-of-the-art detectors.

Results and Discussions

Experimental results demonstrate the effectiveness of PCB-YOLO: it achieves 95.6% mAP50 on the PCB defect dataset, outperforming YOLOv8n by 2.9%. It also reduces parameters by 23.6% (from 3.006 M to 2.302 M), FLOPs by 22.2% (from 8.1 G to 6.3 G), and model weight by 1.3 M. Moreover, ablation studies are finished to evaluate each proposed component contributes positively: the Conv_SWS module boosts recall by 5% and mAP50 by 2.4%; the FBC2f module reduces parameters by 23.5% with a moderate mAP50 trade-off. the NWD-IoU loss increases mAP50 by 2.3% and mAP50-95 by 2.8%. The full integration of all three modules achieves the best balance of accuracy and efficiency. Comparative evaluations demonstrate that PCB-YOLO outperforms Faster R-CNN, YOLOv5n, YOLOv10n, YOLOv11n, YOLOv12n, and the recently introduced lightweight PIC2f-YOLO in terms of both accuracy and model size. Additionally, it achieves a real-time inference speed of 160 f/s on GPU and 50 f/s on CPU, meeting the throughput demands of industrial applications. Moreover, PCB-YOLO also demonstrates strong generalization on external datasets, attaining mAP50 scores of 98.1% on the industrial aluminum surface defects dataset and 93.6% on the PKU-Market-PCB dataset, thereby verifying its robustness and transferability. Visual comparisons reveal that PCB-YOLO substantially decreases the rates of missed detections and false alarms relative to YOLOv8n, especially in the case of minor defects like shorts and mouse bites. In summary, the results validate that the Conv_SWS module effectively preserves small-object features during downsampling, the FBC2f module maintains representational capacity while reducing computation, and the NWD-IoU loss stabilizes training and improves localization for miniature targets.

Conclusions

This paper proposes PCB-YOLO, a lightweight and accurate detection model designed for tiny PCB defects. The approach integrates grayscale transformation and adaptive slicing during preprocessing to enhance input data quality. It introduces the Conv_SWS module to improve small-object feature retention and designs the parameter-efficient FBC2f module. Additionally, the NWD-IoU loss is adopted to refine the regression of minor defects. Together, these contributions address key challenges in PCB inspection, including limited sample availability, background variation, small defect size, and the accuracy–complexity trade-off. Experimental results show that PCB-YOLO achieves higher detection accuracy with fewer parameters and faster inference than current leading models. It also demonstrates strong generalization across diverse industrial defect datasets. The framework offers a practical, deployable solution for real-time, high-precision PCB quality inspection in manufacturing. Future work may explore further architectural optimizations and extend the method to other micro-defect detection domains.

PCB detection  /  tiny object detection  /  YOLOv8n  /  evaluation metrics  /  attention mechanism
邓文斌, 郭怡希, 林源强, 鹿业波, 杨俊. PCB-YOLO:改进YOLOv8n的PCB微小缺陷轻量化检测方法. 光电工程, 2026 , 53 (2) : 250252 -1 . DOI: 10.12086/oee.2026.250252
Wenbin Deng, Yixi Guo, Yuanqiang Lin, Yebo Lu, Jun Yang. PCB-YOLO: an enhanced YOLOv8n for lightweight detection of tiny defects in PCB[J]. Opto-Electronic Engineering, 2026 , 53 (2) : 250252 -1 . DOI: 10.12086/oee.2026.250252
为了适应市场对复杂电子电路日益增长的需求,印制电路板(printed circuit board, PCB)正不断向高精度、多层化与微型化演进,在工业化大规模生产过程中难免导致产品缺陷[1]。PCB上的微小缺陷可能导致整个电子设备的失效,因此对PCB进行高精度的缺陷检测至关重要。由于PCB制造工艺复杂,缺陷尺寸越来越小,这给检测带来了新的挑战。如何准确检测PCB上的微小缺陷,成为当前PCB制造过程研究的热点问题。
在PCB生产过程中,常见的缺陷主要有鼠咬、开路、短路、毛刺和伪铜。这些缺陷的特点如下:1) PCB电路设计多种多样,导致PCB缺陷的背景差异很大;2)缺陷面积小,颜色与背景相似,增大了检测难度;3)虽然存在多种缺陷,但它们的特征是相似的,这使得检测更加困难,容易出现误检和漏检。这些特性可能会导致PCB表面缺陷检测存在以下问题:1)复杂的纹理导致前景和背景区分困难;2)微小的尺寸导致检测困难。由于缺陷目标小、种类多、区分难等因素,传统检测技术已经不能满足实际生产需求[2]。
为应对上述挑战,PCB 缺陷检测技术已从传统人工检测演进至智能算法驱动的自动检测。早期 PCB 缺陷检测依赖人工目视,易受视觉疲劳影响,误检率超10%、效率较低,难以适配大规模生产;X 射线检测虽能穿透多层 PCB,但设备成本超百万元,且难以完成表面缺陷检测。自动光学检测 (Automated optical inspection,AOI)[3]是当前工业主流检测技术,其中基于传统图像处理的 AOI 通过手工设计特征实现缺陷识别,Tsai等人[4]提出了一种全局傅里叶图像重建方法来检测和定位PCB表面缺陷,该方法可以适应平移变化和光照变换环境。Hassanin等人[5]提出了一种基于SURF特征和形态学运算的实时PCB自动缺陷检测方法,可以准确地确定缺陷位置和缺陷类型。这类方法的核心局限在于特征依赖人工设计、鲁棒性差,无法适应 PCB多样化底纹与微小缺陷的检测需求。
为突破人工设计特征的限制,研究者引入机器学习算法实现特征自动提取,Tsai 等人[6]将颜色与结构特征输入SVM分类器,提升了针孔、露铜缺陷的检测精度,但SVM对高维特征的处理能力有限,当PCB 背景存在噪声时准确率下降15% 以上;Lu 等人[7]提出贝叶斯特征融合框架,改善了光照不均场景的检测效果,但模型仍需依赖边缘、纹理等手工设计的基础特征,对微小缺陷的特征表征能力不足,这类方法虽比传统图像处理更具适应性,但未摆脱对人工特征的依赖,且模型泛化能力受限于训练样本分布,面对PCB 工业场景中 “小样本、多缺陷类型” 的特点时,难以满足检测精度要求。
深度学习通过端到端特征学习,显著提升了 PCB 缺陷检测的鲁棒性,按网络结构可分为两阶段模型与单阶段模型。两阶段模型 (如 Faster R-CNN)通过 “区域提议-分类回归” 实现高精度检测,在 PCB 小缺陷检测中得到广泛探索,Hu等人[8]使用GA-RPN (region proposal network by guided anchoring)改进Faster RCNN算法,从而提高了PCB中小缺陷的检测能力。同时,Zeng等人[9]提出了一种基于非对称平衡特征金字塔网络 (asymmetric balanced feature pyramid network, ABFPN)的增强多尺度特征融合算法,实现了有效的小目标检测。孙志超等人[10]通过引入可变形残差卷积替换VGG16网络改进Faster RCNN,并结合IoU Loss等组合优化检测网络,实现了对PCB缺陷图像数据的检测精度和速度。两阶段模型的核心瓶颈在于参数量大、推理慢,难以适配边缘设备的轻量化需求,因此单阶段模型 (尤其是 YOLO 系列)成为近年的研究热点。
YOLO系列模型[11-14]将检测任务整合为单阶段回归,兼顾速度与精度,成为PCB 缺陷检测的主流选择,但现有改进仍存在三大待解决问题:1)小目标特征提取不足。何国忠和梁宇[15]在YOLOv4中引入长距离全局注意力,增强浅层特征信息,但未针对PCB微小缺陷的“特征稀疏性”优化,当缺陷尺寸小于8 pixel×8 pixel时,注意力权重易被背景噪声干扰。Xia等人[16]提出了结合全局上下文注意力和双向加权特征金字塔网络(bi-directional weighted feature pyramid network, BiFPN)特征融合的YOLO模型,提升了多尺度特征融合效果,但未改进下采样过程,小目标特征在池化操作中仍会丢失。2)轻量化与精度平衡不足。李振华等人[17]采用轻量化GhostNet网络作为算法模型的特征提取网络,减少YOLOv5算法参数量,但mAP50下降。Bhattacharya等人[18]用Transformer替换YOLOv5骨干网络,虽然保持了精度,但参数量反增超20%。Lei等人[19]通过张量分解简化卷积,轻量化效果显著,但特征交互能力减弱,对复杂底纹PCB的鲁棒性不足。3)损失函数适配性差。现有YOLO模型多采用CIoU损失,但PCB微小缺陷的边界框标注误差会导致CIoU损失值剧烈波动,影响收敛稳定性。Wise-IoU、Shape-IoU虽优化了回归策略,但未考虑小目标“面积占比低、重叠率低”的特点,当预测框与真实框无重叠时,模型仍无法提供有效梯度。综上,现有PCB缺陷检测技术仍存在三大核心问题:微小目标特征易丢失、模型轻量化与精度难以平衡、损失函数对小目标适配性差。这些问题直接导致检测效率低、部署成本高,无法满足工业需求。
针对以上问题,本文以YOLOv8n算法为基线模型,提出了PCB-YOLO的轻量化PCB缺陷检测算法,解决PCB底纹颜色差异大、缺陷尺寸小导致误检和漏检及模型复杂不易部署到边缘处理器的问题。首先对原始PCB数据集进行预处理,获得符合模型需要的数据集;其次,结合SimAM注意力机制和合并切片操作设计了Conv_SWS模块,将其用于网络的卷积模块中,代替传统的下采样方式,增强了PCB中微小目标缺陷的特征提取并提高了检测准确性;然后通过引入部分卷积来对关键特征通道进行综合计算,设计了FBC2f模块,用于替换YOLOv8中原有的C2f模块,从而增强特征提取能力并降低模型参数量和计算复杂度;最后通过改进损失函数来加速模型收敛,进一步提高PCB-YOLO模型对PCB微小缺陷目标的检测能力。
本文使用的原始数据集来源于华为云竞赛平台 (https://competition.huaweicloud.com/information/ 1000042047/circumstance)。原始数据集包含68张原始PCB表面缺陷图像,图片尺寸为3072 pixel×2048 pixel,每张图片包含一个及以上的缺陷目标。这些目标囊括了PCB生产流程中可能出现的多种缺陷类型,包含鼠咬(mouse_bite)、开路(open_circuit)、短路(short)、毛刺(spur)以及伪铜(spurious_copper) 5类缺陷,其缺陷类型如图1所示。
在PCB生产过程中,存在多种颜色的底纹,而数据集给出的每种样本数量又少,为了增强模型识别的可靠性,降低不同底纹颜色对缺陷识别的干扰,首先对样本数据进行了灰度化处理,将不同颜色底纹的PCB板RGB图像使用式(1)统一转化为灰度图[20],其中:R、G、B分别表示红色、绿色、蓝色。图2展示了不同底纹背景的PCB转换为灰度图以后的效果,从结果可以看到这样可以减少不同背景颜色对缺陷识别的干扰。
$ Gray = R \times 0.299 + G \times 0.587 + B \times 0.114\;. $
由于采集的PCB图像分辨率较高,且待检测的缺陷目标往往尺寸较小,而YOLOv8模型默认的输入图像尺寸为640 pixel×640 pixel,若直接将原始图像强制缩放到该尺寸,不仅会在训练阶段占用大量显存资源,对计算设备提出较高要求。更严重的是,这种处理方式可能导致小尺度缺陷的关键特征在尺寸变换过程中被过度压缩而消失,最终影响模型的识别效果。因此,为缓解高分辨率PCB图像在输入YOLOv8模型时因尺寸压缩而可能造成的微小缺陷漏检问题,本文引入图像分块预处理方法,对原始图像进行切分处理。将原图像连续分割为多个尺寸为640 pixel×480 pixel的子图像,每个子图像拥有100 pixel的重叠区域,避免因缺陷位于滑动窗口边缘而造成漏检,本方法在图像分割处理过程中,依据原始标签文件所提供的缺陷区域坐标,判断当前滑动窗口所覆盖范围内是否包含缺陷目标。通过该机制,可有效筛选出含有缺陷的PCB图像切片,并剔除仅包含背景信息的无效切片,从而提升检测效率与准确性。
经过图像预处理后,针对PCB缺陷数据集中存在的样本分布不平衡及数量有限的问题,本文采用水平镜像翻转、竖直镜像翻转、添加均值为0、标准差为0.1的高斯噪声等多种数据增强方法对原始数据集进行扩充。经增强后,数据集规模增至5172张图像,每幅图像均包含至少一个缺陷,缺陷总数达到6282个。图3展示了数据集中缺陷的高宽比分布情况,从分析结果可以看出,缺陷的高宽比数值覆盖范围较广,属于多尺度特征目标,给检测带来了较高的难度。
图4统计了数据集中所有缺陷的类别与尺寸分布及数量。由图可知,缺陷尺寸主要集中在32 pixel×32 pixel以内,所占比例高达98.3%,符合小目标检测的典型范畴[21]。此外,还存在546个尺寸小于8 pixel×8 pixel像素的极微小缺陷,占比约8.7%,图5展示了多个小于8 pixel×8 pixel的微小缺陷。这类缺陷在多次下采样操作中极易丢失细节信息,导致其在YOLOv8模型中的检测难度显著增加。由于较低的区域覆盖率和较少的特征点,微小缺陷和背景混合在一起形成低对比度,这使得微小缺陷的检测具有极大的挑战性。将扩增后的PCB数据集采用8∶1∶1的比例随机划分为训练集、验证集和测试集,确保各数据集均囊括了多种类型及尺度的缺陷样本,以保证数据的多样性。
YOLOv8模型是Ultralytics公司基于YOLOv5进行优化改进的,在准确率和速度方面都得到了较大提升。YOLOv8的网络结构主要由三个部分组成:提取特征的主干网络、执行特征融合的颈部网络和目标定位及分类的预测头网络,结构细节如图6所示。YOLOv8的骨干网络采用C2f模块[22]替换C3模块,并通过Bottleneck模块(m表示BottleNeck模块的数量)和SPPF模块提升特征提取能力,同时降低计算复杂度和延迟[23]。颈部网络部分使用PAN-FAN结构进行多尺度特征融合,删除上采样层前的卷积层,对模型进行压缩。预测头使用目前主流的解耦头结构,将分类和检测头分离,并使用Anchor Free检测方法[24],不再依赖先验锚框,使模型能够针对不同任务进行优化。
为解决PCB缺陷检测中微小目标识别精度低、漏检/误检率高的问题,以及完成对实时性要求的双重挑战,需要保证模型的轻量化和低参数量,本文提出了一种基于YOLOv8n改进的PCB-YOLO模型,所提出的模型结构如图7所示。
PCB-YOLO的主干网络包含普通卷积 (Conv)、Conv_SWS卷积模块、FBC2f和SPPF模块。Conv模块通过Conv2d、BatchNorm2d和SiLU激活函数对输入图像进行通道数调整与预处理;Conv_SWS模块则融合卷积与切片操作,有效捕获不同尺度目标的关键特征。FBC2f模块利用PConv有效降低网络的计算量和参数量,并通过FasterNet增加特征交互;SPPF模块进一步提升特征提取能力,同时降低计算复杂度,有助于PCB-YOLO 模型的小目标检测能力。
PCB-YOLO 的颈部网络包含Conv_SWS和 FBC2f模块,Conv_SWS模块通过融合不同大小特征图尺寸,实现多尺度特征融合;FBC2f模块采用部分卷积结构构建特征提取单元,在提升局部特征捕获效率的基础上有效降低了计算复杂度,从而进一步强化了PCB-YOLO模型的轻量化性能。
SimAM (Simple attention mechanism)[25]是一种三维自注意力模块,能够同时整合空间与通道信息。该模块以轻量化著称,可在不引入额外参数的情况下,有效引导网络聚焦于输入特征中的关键区域,从而增强模型的特征提取与表示能力。与现有的1D或2D注意力模块不同,SimAM直接在特征图中的每个神经元上运行。它使用能量函数计算3D特征权重,其中权重不仅考虑通道维度,还考虑空间维度。这使得能够更精确地关注特征图中的重要信息。SimAM注意力机制的集成过程如图8所示。首先,从特征X生成3D注意力权重,然后将该注意力权重集成到特征图中,最后应用扩展操作来增强特征图的表示能力。
对于特征图$ \boldsymbol{X}\in\mathrm{\mathbf{R}}^{C\times H\times W} $(其中H、W为特征图高度和宽度,C为通道数),SimAM的最小能量函数et如式 (2)所示,用来衡量目标神经元t的重要性。最小能量值et的大小反映了缺陷特征中神经元t在对应维度上的独特性。该值越小,表明该神经元与同维度其他神经元之间的差异性越显著,其重要性也随之提升。因此,神经元的重要性亦可通过1/et来量化。
$ {e_t} = \dfrac{{4 ({\sigma ^2} + \lambda )}}{{{{ (t - u)}^2} + 2{\sigma ^2} + 2\lambda }}\;, $
式中:t表示输入特征图在单一通道内的目标神经元;$ \lambda $表示正则化系数;u和$ {\sigma ^2} $分别表示通道中所有神经元取值的均值和方差,其表达式分别如式(3)和式(4)所示。
$ u = \dfrac{1}{M}\displaystyle\sum\limits_{i = 1}^M {{x_i}} \;,$
$ {\sigma ^2} = \dfrac{1}{M}\displaystyle\sum\limits_{i = 1}^M { ({x_i}} - u{)^2}\;, $
式中:xi表示目标神经元在特定维度上的全部神经元集合;M=H×W表示该层神经元总数。
最后,依据注意力机制的基本原理,对以上特征信息进行特征增强,获得增强后的输出特征$ \tilde {\boldsymbol{X}} $,如式 (5)所示。
$ \tilde {\boldsymbol{X}} = sigmoid\left (\dfrac{1}{E}\right) \odot {\boldsymbol{X}} \;,$
式中:$ \odot $表示按元素相乘;E表示所有通道和空间维度中et的集合,通过sigmoid对其数值范围进行约束;X表示图像的输入特征。
然而,当SimAM计算整个特征图上像素差异的加权平均值时,它可能会忽略小目标,从而减弱其增强小目标的能力。PCB图像中的缺陷区域通常占比较小,其在整体图像中的特征与背景均值相近,使得SimAM注意力机制在计算权重时难以对其产生显著的增强效果,因此该机制在处理此类小目标缺陷时表现相对有限。为解决这个问题,本文通过引入切片操作来优化SimAM,从而开发SWS (SimAM with slicing)模块。当特征图被划分为多个局部区块时,大目标可能会因分割而损失部分关键特征,进而引起其权重下降;但在后续的特征整合过程中,这些目标仍能够保持较高的可辨识度,同时其显著特征也得到进一步强化;另一方面,经过切片操作的小目标会获得更多的权重,从而增强其特征。改进的SWS模块确保了大、小目标都获得增强,从而提高模型的识别能力。切片操作过程如图9所示。
Backbone作为YOLO算法中特征提取的关键部分,为提升模型对PCB中小目标缺陷的检测效果,设计Conv_SWS模块用于捕捉多尺度特征。Conv_SWS在标准卷积中引入SWS模块来增强普通卷积的特征提取能力,且未引入额外的参数量。图10为Conv_SWS模块的结构图,其中:s表示步长,p表示填充。
Conv_SWS模块的工作流程如下:
1)对输入特征图$ \boldsymbol{X}\in\mathrm{\mathbf{R}}^{H\times W\times C} $ 进行批归一化和SiLU激活函数处理,如式 (6)所示。
$ {{\boldsymbol{X}}_{{\rm{bn}}}} = SiLU (BatchNorm2d ({\boldsymbol{X}}))\;. $
2)将标准化后的特征图Xbn划分为K2个局部区块,记为F(i, j),则有
$ {\boldsymbol{F}} (i,j) = {{\boldsymbol{X}}_{{\rm{bn}}}}[i:H:K,j:W:K]\;, $
式中:$ i,j \in [0,K - 1] $。
3)对每个局部区块$ \boldsymbol{X}_{\rm{block,\mathit{i},\mathit{j}}} $独立计算SimAM注意力权重,强化区块内缺陷特征的响应,具体计算过程基于式(2)~(5)。
4)将K2个注意力增强后的局部区块F(i, j)按原分块位置在通道维度上重新拼接为完整特征图$ \boldsymbol{X}_{\rm{merge}}\in\mathrm{\mathbf{R}}^{H\times W\times C} $,拼接顺序与步骤2)的分块顺序一致,确保特征空间位置的连续性。
5)对合并后的特征图卷积进行下采样,实现特征图尺寸减半,同时保持通道数不变。
在PCB缺陷检测中,存在大量尺寸小于32 pixel×32 pixel的微小缺陷。针对此类小目标,传统的3×3卷积核难以充分捕获其完整特征,而采用更大尺度的卷积核则容易丢失细节信息,从而削弱网络对小缺陷特征的判别能力。本文所引入的Conv_SWS卷积模块,在网络中承担下采样功能。相比原有结构,该模块能够显著缓解下采样过程中的信息损耗,同时增强特征之间的交互关系,并有效扩大感受野范围,进而提升网络对多尺度缺陷的检测性能。
YOLOv8主干网络采用的C2f模块,该结构由多个Darknet Bottleneck堆叠而成,在一定程度上增大了模型的复杂性与计算冗余。为缓解该问题,Chen等[26]引入了部分卷积 (partial convolution, PConv),该方法通过仅对输入特征图的部分通道进行卷积处理,显著降低了模型的参数总量与计算复杂度。PConv通过仅对输入通道的一个子集应用标准卷积,同时保持其余通道不变,从而利用特征图的冗余性。图11展示了PConv的工作原理。
假设h、w分别为输入数据的高度和宽度,c为输入通道数,cp为部分输入通道数,k为卷积核大小,标准卷积和部分卷积的计算量如式(8)~(10)所示。
$ FLOP1 = h \times w \times {k^2} \times {c^2}\;, $
$ FLOP2 = h \times w \times {k^2} \times c_{\rm{p}}^2\;, $
$ \dfrac{FLOP1}{FLOP2}=\dfrac{c^2}{c_{\mathrm{p}}^2}\; . $
当cp取为c的数量的通道时,PConv的浮点计算量为标准卷积的1/16。因此,PConv能够有效地提升计算效率和降低算法所需的计算资源。
为进一步提高模型提取特征能力,将PConv组合卷积网络,创建FasterNet模块,如图12所示。为提升模型效率,本文采用FasterNet Block替换C2f模块中原有的BottleNeck模块,同时保持C2f的整体结构不变,由此提出一种快速卷积交叉融合模块 (FasterNet block cross stage partial fusion,FBC2f),其结构如图13所示。具体来说,输入特征张量首先通过1×1卷积进行处理,使输出特征张量通道的数量翻倍,从而增强模型的特征表示能力。在FBC2f模块中,输入特征张量分为两个部分:其中一部分输入至FasterNet block进行特征变换,另一部分则直接保留并与处理后的特征进行拼接。经过分割的特征将在多个级联的FasterNet block中逐步处理,从而有效捕获更深层的特征。集成的PConv减少了冗余计算和内存访问,同时保持了有效的空间特征提取。随后,所有FasterNet Block的输出与先前分割得到的特征张量进行拼接,以提升特征表达的多样性。最后,通过一个卷积层对拼接后的特征进行通道压缩,使其匹配后续处理所需的输出维度。这种改进不仅保持了高FLOPs,还显著降低了FLOPs,有效减少了不必要的计算和内存使用[27]。因此,它减小了模型尺寸并加速了推理时间,更适合轻量化部署。
在PCB缺陷数据集中,缺陷目标通常尺寸较小、形态多样,且标注质量存在不一致现象,例如部分缺陷定位不精确或边界框标注可靠性较低。原方法采用CIoU损失作为边界框回归的损失函数,该函数在IoU基础上引入中心点距离惩罚项,以促进预测框与真实框的中心对齐,从而加快收敛速度。然而,该方法对所有样本无论其标注质量高低,均采用统一的损失计算方式。而IoU对小目标的位置偏差非常敏感,在基于Anchor的检测器中使用时,轻微的位置偏移就会导致标签分配不准,影响检测精度。为了解决这一问题,本文引入了一种基于Wasserstein距离的度量标准NWD (normalized Wasserstein distance)[28]与标准IoU结合,按照一定比例进行加权得到最终的损失函数。由于NWD对目标尺度不敏感,并且在目标没有重叠的情况下也能度量相似性,因此更适合测量小目标。NWD采用二维高斯分布对边界框进行建模,并利用两个高斯分布之间的Wasserstein距离来评估边界框之间的相似度,Wasserstein距离的计算公式为
$ \begin{split} & W_2^2 ({N_{\rm{a}}},{N_{\rm{b}}}) = \\& \left\| {\left ( {{{\left[ {c{x_{\rm{a}}},c{y_{\rm{a}}},\dfrac{{{w_{\rm{a}}}}}{2},\dfrac{{{h_{\rm{a}}}}}{2}} \right]}^{\mathrm{T}}},{{\left[ {c{x_{\rm{b}}},c{y_{\rm{b}}},\dfrac{{{w_{\rm{b}}}}}{2},\dfrac{{{h_{\rm{b}}}}}{2}} \right]}^{\mathrm{T}}}} \right)} \right\|_2^2\end{split} \;,$
式中:Na、Nb分别代表预测框A和真实框B的二维高斯分布; (cxa, cya)和 (cxb, cyb)分别代表预测框A和真实框B的中心点坐标;wa和ha分别代表预测框A的宽度和高度;wb和hb分别代表真实框B的宽度和高度。
对$ W_2^2(N_{\mathrm{a}},N_{\mathrm{b}}) $进行归一化处理,得到NWD为
$ NWD(N_{\mathrm{a}},N_{\mathrm{b}})=\exp\left(-\dfrac{\sqrt{W_2^2(N_{\mathrm{a}},N_{\mathrm{b}})}}{\mathrm{C}}\right)\; , $
式中:C是与数据集相关的常数,根据文献[28]中的研究结果,实验中将C设成数据集中目标的平均绝对尺寸,且在一定范围内,结果都是稳定的。根据图4统计的数据集结果,可设定C的值为16。由于$ NWD(N_{\mathrm{a}},N_{\rm{b}}) $表示预测框A和真实框B之间的归一化Wasserstein距离,因此损失函数可设为
$ {L_{{\mathrm{NWD}}}} = 1 - NWD ({N_{\rm{a}}},{N_{\rm{b}}})\;. $
将NWD和IoU结合,得到改进后的损失函数计算公式为
$ {L_{{\mathrm{NWDIoU}}}} = \alpha \times {L_{{\mathrm{NWD}}}} + (1 - \alpha ) \times (1 - IoU) \;,$
式中:$ \alpha $为权重系数。调节NWD对小目标的优化优先级与IoU对大目标的稳定性。经实验验证,本文最终将$ \alpha $设为0.5。
本文实验使用Win 10系统,CPU是Intel Core i7-12700KF,GPU为NVIDIA GeForce RTX 4060Ti。编程语言是Python 3.9,采用的深度学习框架是PyTorch 1.12.0框架、CUDA版本是CUDA 11.6。在模型训练过程中,优化器选用随机梯度下降 (SGD),初始学习率设为0.001,动量参数设置为0.9,权重衰减系数为0.0005,其余参数保持系统默认配置。输入图像设置为640 pixel×640 pixel,批处理大小(batch size)设置为16,训练轮数(epochs)共计200轮。
为了评估算法的检测性能,实验中使用平均精度(mean average precision,mAP)作为主要评价指标,由精度(precision, P)、召回率(recall, R)和平均精度(AP)计算所得,其计算公式如式(15)~(18)所示。此外,为了评估模型的复杂度和计算成本,本文还采用了模型参数量(Params)、浮点运算量(FLOPs)和模型大小(Weights)等参数来评价模型的轻量化程度。
$ P=\dfrac{TP}{TP+FP}\; , $
$ R=\dfrac{TP}{TP+FN}\; , $
$ AP = \displaystyle\int_0^1 {P (R){\rm{d}}R} \;,$
$ mAP = \dfrac{1}{n}\displaystyle\sum\limits_{i = 1}^n {A{P_i}} \;,$
式中:TP表示正确检测出对应缺陷的预测框的数量;FP表示检测错误的预测框的数量;FN表示未能检测出的预测框的数量;AP表示是P-R曲线和坐标轴包围的面积;mAP表示所有检测类别的AP的平均值;APi表示数据集中特定类别的平均精度;n表示缺陷种类的数量,在本文中,n=5。
采用YOLOv8n与PCB-YOLO方法对增强前后的PCB缺陷数据集进行实验,结果如图14所示 (“+”代表增强后的数据集)。实验数据显示,YOLOv8n在原始数据集上的mAP50和mAP50-95分别为87.0%与44.4%;而在数据增强后,这两项指标分别提升至92.7%和59.6%,增幅达到5.7%和15.2%。PCB-YOLO在增强前实现了89.3%的mAP50与44.6%的mAP50-95,增强后则进一步提高到95.6%和62.3%,分别上升了6.3%和17.7%。由此可见,数据增强显著提升了两种模型的检测性能,尤其对于本文提出的PCB-YOLO方法,在图像数量较多的数据集上展现出更优异的学习能力。
本实验基于YOLOv8n模型,采用了两种方法来减少模型的参数量和计算量。首先,采用主流轻量级特征提取网络,如GhostNetv2[29]、VanillaNet[30]用以代替原始模型的标准卷积模块,实现模型的轻量化。其次,采用本文设计的FBC2f来代替原始模型的C2f模块,实验结果如表1所示。
实验结果表明,当引入VanillaNet作为特征提取网络时,模型实现了最低的参数量、浮点运算量和模型权重,分别是原始模型的60.8 %、69.1 % 和62.7 %,但mAP50和mAP50-95显著下降。当使用FBC2f作为轻量化方法时,与GhostNet和VanillaNet相比,mAP50分别增加了0.2%和1.4%;与原始模型相比,其参数量、浮点运算量和模型权重分别减少了23.5%、22.2%和22.0%。因此,FBC2f模块更好地平衡了模型轻量化程度和检测精度。
为了验证改进的轻量级注意力机制(SimAM)的有效性,对改进YOLOv8n中使用的SimAM注意力模块进行了替换实验。将SimAM分别替换为ECA (efficient-channel-attention)[31]、DAT (deformable attention)[32]注意力机制模块及最初的SimAM进行了对比分析,结果详见表2。
实验结果表明,SimAM注意力机制和ECA注意力机制作比较时,在相似的参数量、浮点运算量和模型权重下,SimAM的mAP50提高了0.6%。此外,与DAT注意力机制作比较,SimAM将参数量和模型权重分别减少1.149 M和1.800 M。与原始SimAM模块相比,改进的SWS模块将mAP50和mAP50-95分别提高了1.3%和1.8%,达到了95.1%和62.1%。这表明在模型中融合SimAM注意力机制和切分方法,更适合多尺度缺陷检测任务。
为了验证Wasserstein 距离损失函数的有效性,在Wasserstein 距离损失函数 (NWD-IoU)和其他损失函数之间进行了比较实验,包括CIoU (complete intersection over union)、EIoU (efficient intersection over union)[33]、Wise-IoU[34]和Shape-IoU[35],实验结果如表3所示。
实验结果表明,与原始CIoU损失函数相比,NWD-IoU损失的精度、召回率和mAP50分别提高了3.9%、3.8%、3.4%。与其他损失函数相比,NWD-IoU实现了最高的精度、召回率和mAP50和mAP50-95。此外在mAP50检测性能方面,相比EIoU、Wise-IoU和Shape-IoU 分别提高了4.8%、1.7%和1.1%,这说明NWD-IoU损失有利于增强微小目标的检测性能。
为验证损失函数中权重系数α的最优性,在PCB 缺陷数据集上设计灵敏度分析实验,将α分别设置为 0.1、0.3、0.5、0.7、0.9,实验结果如表4所示。
由表4可知,当α<0.5时,NWD 的权重不足,IoU的主导作用导致微小缺陷的回归梯度不足,召回率、mAP50和mAP50-95随α的增大而上升,表明NWD的加入显著改善了小目标的漏检问题。当α>0.5时,NWD 的权重过高,虽对小目标的梯度引导增强,但IoU 对大目标的约束减弱,导致中等尺寸缺陷的边界框定位精度下降,精度、mAP50和mAP50-95随α的增大而下降,表明过度依赖 NWD 会牺牲大目标的检测稳定性。当α=0.5时,模型取得最优综合性能:精度达到96.3%,召回率达到91.6%,mAP50达到95.6%,以及mAP50-95达到62.3%,证明该取值能较好平衡小目标与大目标的回归需求。
为了验证每个改进模块在PCB缺陷检测中的有效性,以YOLOv8n为基线模型,在构建的PCB缺陷数据集上进行6组消融实验。具体地,M1实验将主干网络和颈部网络中的C2f 模块统一替换为 FBC2f 模块;M2实验则将主干和颈部网络中的下采样模块调整为Conv_SWS模块;M3实验将损失函数替换为NWD-IoU。通过测试改进模块的不同组合来进行消融实验,验证每项改进的有效性,实验结果见表5。
在主干和颈部网络融入FBC2f模块后,模型在mAP50指标上较基线模型YOLOv8n降低了2.2%,但参数量减少了0.705 M,浮点运算量减少了1.8 G。使用Conv_SWS模块后的方法与基线相比,召回率提高了5%,达到92.8%,指标排名第一,mAP50和mAP50-95分别提高了2.4%和2.5%,参数量、浮点运算量和模型权重保持不变。采用NWD-IoU损失函数的方法与基线相比,在参数量、浮点运算量和模型权重保持不变的情况下,mAP50和mAP50-95分别提高了2.3%和2.8%。结合FBC2f和Conv_SWS的方法与使用FBC2f的方法相比,mAP50和mAP50-95分别提升1.7%和2.8%,与基线模型相比,参数量、浮点运算量和模型权重都有较大幅度下降。融合FBC2f、Conv_SWS和NWD_IoU的方法与基线相比mAP50提升2.9%,指标排名第一,浮点运算量减少1.8 G,模型权重减少1.3 M,指标排名第一。
引入FBC2f模块后的模型与基线相比,mAP50下降至90.5%,但参数量和浮点运算量分别下降了23.5%和22.2%。这表明在网络中使用FasterNet模块能显著降低模型复杂度,增加模型轻量化程度。在此基础上进一步引入Conv_SWS模块,mAP50提高至92.2%,mAP50-95提高至51.1%,并保持了模型的轻量化程度,说明引入的多特征融合注意力机制能有效提取丰富的上下文信息,进而提升模型的多尺度缺陷识别性能。在此基础上进一步引入NWD_IoU损失函数后,mAP50和mAP50-95分别提升至95.6%和62.3%,表明它对于微小目标的检测有较好的提升作用,同时又不影响计算速度。消融实验印证了PCB-YOLO模型的有效性,其在检测精度、推理效率与模型复杂度之间取得了良好的均衡性,这一优势为其实际部署与应用奠定了坚实基础。
为了进一步验证所提算法的优越性和有效性,使用Faster R-CNN[36]、YOLOv5n、YOLOv8n、YOLOv10n[37]、YOLOv11n、YOLOv12[38]、最近发表的轻量化模型PIC2f-YOLO[39]和所提出的模型进行了比较实验,实验结果如表6所示。基于CPU和GPU的帧率(frame per second, FPS)均为将模型转化为.onnx格式后在实验过程中用计算机进行部署后测量所得。
实验结果表明,Faster R-CNN作为经典的二阶段检测算法,其检测精度最低,推理速度最慢,模型权重最大,不适合实时应用。在YOLO系列模型中,PCB-YOLO方法与YOLOv5n、YOLOv10n、YOLOv11n和YOLOv8n等模型相比,其在检测精度、召回率、mAP50和mAP50-95等指标均排名第一,综合表现优于其它模型。与原始的YOLOv8n算法相比,PCB-YOLO算法在精确率、召回率、mAP50和mAP50-95方面分别提升了2.9%、3.8%、2.9%和2.7%,在参数量、浮点运算量和模型权重指标分别下降了0.71 M、1.8 G和1.3 M。对于FPS指标,虽然YOLOv10和YOLOv8n分别在CPU和GPU上获得了最快的推理速度,但其检测精度比改进后的模型下降较多,而改进后的PCB-YOLO模型在GPU上的推理速度达到了160 f/s,在CPU上也达到了50 f/s,完全满足实际工业化应用需求,因此,该模型符合轻量化且高精度要求,使其更适合实时检测需求。
对PCB缺陷数据集的检测效果进行比较,展示了YOLOv8n与本文提出的方法在性能上的差异。如图15(b)所示,YOLOv8n 在短路缺陷中存在漏检现象,其在工厂对PCB高质量生产的实际应用需求中可能难以胜任。相比之下,改进后的PCB-YOLO 方法能够更精确地识别出有效缺陷区域,显著减少了漏检情况。具体而言,在鼠咬、开路、短路、毛刺和伪铜等缺陷类型上,PCB-YOLO 相比原始YOLOv8n表现出更高的检测置信度。因此本文提出的检测方法能更好避免漏检和误检,具有更好的缺陷检测性能。
为验证PCB-YOLO模型的泛化性能,将基线模型、轻量化模型PIC2f-YOLO和本文算法在百度飞桨平台上公开的铝片表面工业缺陷检测数据集[40]和北京大学智能机器人开放实验室的开源PKU-Market-PCB缺陷数据集[41]上进行实验对比。其中铝片表面工业缺陷检测数据集共有412张图片,有针孔、擦伤、脏污、褶皱4种缺陷类型,包含小中大型不同尺寸的缺陷类型。PKU-Market-PCB缺陷数据集包含693张图片,共6种缺陷类型,且以中小型缺陷为主,具有很大挑战性。实验结果如表7和表8所示,在铝片表面工业缺陷数据集上,PCB-YOLO模型的mAP50达到 98.1%, 相比基线模型和PIC2f-YOLO模型分别提高了0.8%和3.1%,同时模型参数量分别减少了0.71 M和0.30 M。在PKU-Market-PCB数据集上,其mAP50也提升至93.6%,较基线算法和PIC2f-YOLO方法分别提高了2.7%和1.2%,参数量同样分别减少0.71 M和0.30 M。
上述结果说明,PCB-YOLO模型在不同类型和缺陷特点的数据集上均表现出良好的检测性能与稳定的泛化能力。
现有PCB缺陷检测中因数据样本少、底纹差异大、相机拍摄图像尺寸大,导致识别准确率不高、误检与漏检发生频次较高,以及处理速度慢等局限性。为此,本文设计了PCB-YOLO模型,能显著提升检测性能,有效应对以上挑战。通过图像灰度化和切片操作,显著提高微小目标缺陷的检测精度;将改进的SimAM注意力机制集成到主干卷积模块中,优化特征间的交互和整合,进一步提高模型对不同尺寸大小的目标感知能力;设计的快速卷积交叉融合模块能在有效提取目标缺陷特征的同时,大幅降低模型的计算量,达到轻量化的目的;最后通过引入NWD作为回归损失函数替换CIoU损失函数,提升模型的效率和准确性。实验结果表明,PCB-YOLO模型有效地应对了平衡模型检测精度和轻量级设计的挑战,并且具有较好的泛化性能。

参考文献 引证文献
排序方式:
1
吴一全, 赵朗月, 苑玉彬, 等. 基于机器视觉的PCB缺陷检测算法研究现状及展望[J]. 仪器仪表学报, 2022, 43(8): 1−17.
Wu Y Q, Zhao L Y, Yuan Y B, et al. Research status and the prospect of PCB defect detection algorithm based on machine vision[J]. Chin J Sci Instrum, 2022, 43(8): 1−17.
2
董静毅, 吕文涛, 包晓敏, 等. 基于机器视觉的PCB板表面缺陷检测方法研究进展[J]. 浙江理工大学学报, 2021, 45(3): 379−389.
Dong J Y, Lü W T, Bao X M, et al. Research progress of the PCB surface defect detection method based on machine vision[J]. J Zhejiang Sci-Tech Univ, 2021, 45(3): 379−389.
3
Ebayyeh A A R M A, Mousavi A. A review and analysis of automatic optical inspection and quality monitoring methods in electronics industry[J]. IEEE Access, 2020, 8: 183192−183271.
4
Tsai D M, Huang C K. Defect detection in electronic surfaces using template-based Fourier image reconstruction[J]. IEEE Trans Compon Package Manuf Technol, 2019, 9(1): 163−172.
5
Hassanin A A I M, Abd El-Samie F E, El Banby G M. A real-time approach for automatic defect detection from PCBs based on SURF features and morphological operations[J]. Multimed Tools Appl, 2019, 78(24): 34437−34457.
6
Tsai D M, Lin B T. Defect detection of gold-plated surfaces on PCBs using entropy measures[J]. Int J Adv Manuf Technol, 2002, 20(6): 420−428.
7
Lu Z S, He Q Q, Xiang X G, et al. Defect detection of PCB based on Bayes feature fusion[J]. J Eng, 2018, 2018(16): 1741−1745.
8
Hu B, Wang J H. Detection of PCB surface defects with improved faster-RCNN and feature pyramid network[J]. IEEE Access, 2020, 8: 108335−108345.
9
Zeng N Y, Wu P S, Wang Z D, et al. A small-sized object detection oriented multi-scale feature fusion approach with application to defect detection[J]. IEEE Trans Instrum Meas, 2022, 71: 3507014.
10
孙志超, 王博, 张晓玲. 基于可变形残差卷积与伸缩式特征金字塔算法的PCB缺陷检测[J]. 电讯技术, 2023, 63(6): 798−805.
Sun Z C, Wang B, Zhang X L. PCB defect detection based on deformable residual convolution and scalable feature pyramid algorithm[J]. Telecommun Eng, 2023, 63(6): 798−805.
11
Redmon J, Farhadi A. YOLOv3: an incremental improvement[Z]. arXiv: 1804.02767, 2018. https://doi.org/10.48550/arXiv.1804.02767.
12
Ultralytics. YOLOv5[EB/OL]. (2025-08-25) https://github.com/ultralytics/YOLOv5.
13
Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464–7475. https://doi.org/10.1109/CVPR52729.2023.00721.
14
Ultralytics[EB/OL]. (2025-08-25) GitHub, 2023. https://github.com/ultralytics/ultralytics.
15
何国忠, 梁宇. 基于卷积神经网络的PCB缺陷检测[J]. 图学学报, 2022, 43(1): 21−27.
He G Z, Liang Y. PCB defect detection based on convolutional neural network[J]. J Graphics, 2022, 43(1): 21−27.
16
Xia K W, Lv Z L, Liu K. et al. Global contextual attention augmented YOLO with ConvMixer prediction heads for PCB surface defect detection[J]. Sci Rep, 2023, 13(1): 9805.
17
李振华, 张雷. 改进YOLOv5的轻量级PCB缺陷检测算法[J]. 无线电工程, 2023, 53(6): 1342−1350.
Li Z H, Zhang L. Lightweight PCB defect detection algorithm of Improved YOLOv5[J]. Radio Eng, 2023, 53(6): 1342−1350.
18
Bhattacharya A, Cloutier S G. End-to-end deep learning framework for printed circuit board manufacturing defect classification[J]. Sci Rep, 2022, 12(1): 12559.
19
Lei L, Li H X, Yang H D. Reliable and lightweight adaptive convolution network for PCB surface defect detection[J]. IEEE Trans Instrum Meas, 2024, 73: 2003208.
20
Gonzalez R C, Wood R E. 数字图像处理[M]. 阮秋琦, 译. 4版. 北京: 电子工业出版社, 2020.
Gonzalez R C, Wood R E. Digital Image Processing[M]. Ruan Q Q, trans. 4th ed. Beijing: Electronics Industry Press, 2020.
21
Tong K, Wu Y Q, Zhou F. Recent advances in small object detection based on deep learning: a review[J]. Image Vision Comput, 2020, 97: 103910.
22
Qiu L L, Zhu W G, Li Y G, et al. YOLO-based multi-class target detection for SAR images[C]//Proceedings of 2023 IEEE 7th Information Technology and Mechatronics Engineering Conference (ITOEC), 2023: 1671–1675. https://doi.org/10.1109/ITOEC57671.2023.10292083.
23
Jooshin H K, Nangir M, Seyedarabi H. Inception-YOLO: computational cost and accuracy improvement of the YOLOv5 model based on employing modified CSP, SPPF, and inception modules[J]. IET Image Process, 2024, 18(8): 1985−1999.
24
Zhang S F, Chi C, Yao Y Q, et al. Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 9756–9765. https://doi.org/10.1109/CVPR42600.2020.00978.
25
Yang L X, Zhang R Y, Li L D, et al. SimAM: a simple, parameter-free attention module for convolutional neural networks[C]//Proceedings of the 38th International Conference on Machine Learning (ICML), 2021: 11863–11874.
26
Chen J R, Kao S H, He H, et al. Run, don't walk: chasing higher FLOPS for faster neural networks[C]//Proceedings of 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 12021–12031. https://doi.org/10.1109/CVPR52729.2023.01157.
27
Wang H R, Cheng H, Zhang J T. Faster-PGYOLO: an efficient framework for floating debris detection in inland waters[J]. Vis Comput, 2025, 41(7): 5087−5104.
28
Wang J W, Xu C, Yang W, et al. A normalized gaussian Wasserstein distance for tiny object detection[Z]. arXiv: 2110.13389, 2021. https://doi.org/10.48550/arXiv.2110.13389.
29
Tang Y H, Han K, Guo J Y, et al. GhostNetV2: enhance cheap operation with long-range attention[C]//Proceedings of the 36th Conference on Neural Information Processing Systems (NeurIPS), 2022: 724.
30
Chen H T, Wang Y H, Guo J Y, et al. VanillaNet: the power of minimalism in deep learning[C]//Proceedings of the 37th International Conference on Neural Information Processing Systems, 2023: 308.
31
Wang Q L, Wu B G, Zhu P F, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020: 11531–11539. https://doi.org/10.1109/CVPR42600.2020.01155.
32
Xia Z F, Pan X R, Song S J, et al. Vision transformer with deformable attention[C]//Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022: 4784–4793. https://doi.org/10.1109/CVPR52688.2022.00475.
33
Zhang Y F, Ren W Q, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146−157.
34
Tong Z J, Chen Y H, Xu Z W, et al. Wise-IoU: bounding box regression loss with dynamic focusing mechanism[Z]. arXiv: 2301.10051, 2023. https://doi.org/10.48550/arXiv.2301.10051.
35
Zhang H, Zhang S J. Shape-IoU: more accurate metric considering bounding box shape and scale[Z]. arXiv: 2312.17663, 2023. https://doi.org/10.48550/arXiv.2312.17663.
36
Ren S Q, He K M, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Trans Pattern Anal Mach Intell, 2017, 39(6): 1137−1149.
37
Wang A, Chen H, Liu L H, et al. YOLOv10: real-time end-to-end object detection[C]//Proceedings of the 38th International Conference on Neural Information Processing Systems, 2024: 3429.
38
Tian Y J, Ye Q X, Doermann D. YOLOv12: attention-centric real-time object detectors[Z]. arXiv: 2502.12524, 2025. https://doi.org/10.48550/arXiv.2502.12524.
39
胡依伦, 杨俊, 许聪源, 等. PIC2f-YOLO: 金属表面缺陷检测轻量化方法[J]. 光电工程, 2025, 52(1): 240250.
Hu Y L, Yang J, Xu C Y, et al. PIC2f-YOLO: a lightweight method for the detection of metal surface defects[J]. Opto-Electron Eng, 2025, 52(1): 240250.
40
铝片表面工业缺陷目标检测[EB/OL]. https://aistudio.baidu.com/datasetdetail/135640.
AI Studio. Surface defect detection for aluminum sheets in industrial applications[EB/OL]. (2025-08-25) https://aistudio.baidu.com/datasetdetail/135640. AI Studio.
41
Ding R W, Dai L H, Li G P, et al. TDD-net: a tiny defect detection network for printed circuit boards[J]. CAAI Trans Intell Technol, 2019, 4(2): 110−116.
2026年第53卷第2期
PDF下载
126
51
引用本文
BibTeX
文章信息
doi: 10.12086/oee.2026.250252
  • 接收时间:2025-08-25
  • 首发时间:2026-10-22
  • 出版时间:2026-02-25
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-08-25
  • 修回日期:2025-10-24
  • 录用日期:2025-10-28
基金
作者信息
    1嘉兴大学机械工程学院,浙江 嘉兴 314001
    2嘉兴大学人工智能学院,浙江 嘉兴 314001

通讯作者:

参考文献
分享链接
https://castjournals.cast.org.cn/joweb/oee/CN/10.12086/oee.2026.250252
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏