| 注册
首页|期刊导航|哈尔滨商业大学学报(自然科学版)|ECARA-YOLO:一种改进的印刷电路板缺陷检测算法

ECARA-YOLO:一种改进的印刷电路板缺陷检测算法

张逸 唐贝贝

哈尔滨商业大学学报(自然科学版)2025,Vol.41Issue(1):43-51,9.
哈尔滨商业大学学报(自然科学版)2025,Vol.41Issue(1):43-51,9.

ECARA-YOLO:一种改进的印刷电路板缺陷检测算法

ECARA-YOLO:An improved printed circuit board defect detection algorithm

张逸 1唐贝贝2

作者信息

  • 1. 安徽理工大学 计算机科学与工程学院,安徽 淮南 232001
  • 2. 安徽理工大学 人工智能学院,安徽 淮南 232001
  • 折叠

摘要

Abstract

As the foundation of the electronic information industry,the quality of printed circuit boards hada decisive impact on the subsequent production of electronic products.However,due to the influence of complex background,subtle defects and irregular shape of printed circuit boards,the existing defect detection of printed circuit boards had problems such as missed detection and false alarm.To solve the above problems,an improved printed circuit board defect detection algorithm ECARA-YOLO was proposed.An efficient channel attention mechanism was introduced,and an adaptive local interaction strategy was used to capture spatial information and enhance the ability to extract the features of small target defects.In addition,a feature fusion network ARANeck was designed by combining the efficient channel attention mechanism and the content-aware feature recombination module.In the content-aware feature recombination module,semantic information and content information were combined and sampled,and target focus was realized through efficient attention mechanism,which made small target defect detection more comprehensive and efficient.Experimental results showed that the improved algorithm had a mAP index of 96.2%,which was 7.1%,9.2%,3.9% and 1.9% higher than the four classical models Fast-RCNN,YOLOv4,YOLOv5s and YOLOv8,respectively.Compared with four classical models,ECARA-YOLO had significant advantages in the detection of six types of defects.Compared with the existing methods,the proposed algorithm effectively improved the detection accuracy of small defects of printed circuit boards in complex environments,and had certain reference value in the field of industrial detection research.

关键词

印刷电路板/缺陷检测/YOLO/高效通道注意力机制/内容感知特征重组/特征融合

Key words

printed circuit board/defect detection/YOLO/efficient channel attention/content-aware reassembly of features/feature fusion

分类

信息技术与安全科学

引用本文复制引用

张逸,唐贝贝..ECARA-YOLO:一种改进的印刷电路板缺陷检测算法[J].哈尔滨商业大学学报(自然科学版),2025,41(1):43-51,9.

基金项目

安徽高校协同创新项目(GXXT-2021-006) (GXXT-2021-006)

哈尔滨商业大学学报(自然科学版)

1672-0946

访问量0
|
下载量0
段落导航相关论文