Scholarlink

<p>To ensure the quality and reliability of electronic devices, it is important to detect surface defects in Printed Circuit Boards (PCBs). Automated PCB defects detection in images is generally associated with various challenges, including, extremely small size of the defects, lack of contextual information, and wide variations in sizes of defects. To address these challenges, we present a novel framework that mainly consists of two main modules, namely, the Feature Aggregation Block (FAB) and the Contextual Information Integration (CII) module. FAB effectively handles scale variation in defects by combining features from different layers, allowing the model to detect defects of various sizes. On the other hand, the CII module enhances contextual information by incorporating dilated convolutions which enables the model to precisely identify small defects with a broader context. To evaluate the effectiveness of the framework, we perform extensive experiments on two publicly available benchmark datasets. Through both qualitative and quantitative assessments, we demonstrate that the proposed approach outperforms other benchmark methods, establishing its effectiveness in identifying various defects within complex PCBs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep learning framework for enhanced PCB surface defect detection leveraging multi-scale feature aggregation and contextual information

  • 期刊论文

  • 发表日期：

    2026年2月13日

查看全文
  • Sultan Daud Khan,
  • Saleh Basalamah

摘要

To ensure the quality and reliability of electronic devices, it is important to detect surface defects in Printed Circuit Boards (PCBs). Automated PCB defects detection in images is generally associated with various challenges, including, extremely small size of the defects, lack of contextual information, and wide variations in sizes of defects. To address these challenges, we present a novel framework that mainly consists of two main modules, namely, the Feature Aggregation Block (FAB) and the Contextual Information Integration (CII) module. FAB effectively handles scale variation in defects by combining features from different layers, allowing the model to detect defects of various sizes. On the other hand, the CII module enhances contextual information by incorporating dilated convolutions which enables the model to precisely identify small defects with a broader context. To evaluate the effectiveness of the framework, we perform extensive experiments on two publicly available benchmark datasets. Through both qualitative and quantitative assessments, we demonstrate that the proposed approach outperforms other benchmark methods, establishing its effectiveness in identifying various defects within complex PCBs.