Scholarlink

<p>With the rapid development of the electronics manufacturing industry, the defect detection of printed circuit boards (PCBs) plays a crucial role in ensuring product quality. This paper presents a lightweight PCB defect detection model named LAP-Net. This model mainly overcomes the difficulties of parameter redundancy and low detection accuracy in existing methods. Specifically, while maintaining the overall framework of YOLOv8n, LAP-Net incorporates an enhanced ShuffleNetV2 structure in its backbone for efficient feature extraction. Furthermore, a novel method using dual convolution attention mechanism is introduced to effectively enhance the accuracy of small-sized PCB defect detection. Meanwhile, introducing a lightweight feature detection head eliminates redundant parameters and thus reduces the overall complexity of the algorithm. Massive experiments conducted on the public PCB dataset demonstrate the superiority of the presented LAP-Net model, while the balance between algorithm complexity and detection precision outperforms other algorithms. Compared with the baseline algorithm YOLOv8n, the LAP-Net reduces parameters by 11.7% and FLOPs by 22.2%, while the mAP50 improves by 3.2%. Therefore, the proposed LAP-Net is validated as a reliable, competitive, and lightweight PCB defect detection model.</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

LAP-Net: a lightweight PCB defect detection network combined with attention mechanisms

  • 期刊论文

  • 发表日期：

    2025年8月21日

查看全文
  • Ziqiang Li,
  • Qing Ai,
  • Ende Peng,
  • Shaoyu Mao,
  • Tao Han

摘要

With the rapid development of the electronics manufacturing industry, the defect detection of printed circuit boards (PCBs) plays a crucial role in ensuring product quality. This paper presents a lightweight PCB defect detection model named LAP-Net. This model mainly overcomes the difficulties of parameter redundancy and low detection accuracy in existing methods. Specifically, while maintaining the overall framework of YOLOv8n, LAP-Net incorporates an enhanced ShuffleNetV2 structure in its backbone for efficient feature extraction. Furthermore, a novel method using dual convolution attention mechanism is introduced to effectively enhance the accuracy of small-sized PCB defect detection. Meanwhile, introducing a lightweight feature detection head eliminates redundant parameters and thus reduces the overall complexity of the algorithm. Massive experiments conducted on the public PCB dataset demonstrate the superiority of the presented LAP-Net model, while the balance between algorithm complexity and detection precision outperforms other algorithms. Compared with the baseline algorithm YOLOv8n, the LAP-Net reduces parameters by 11.7% and FLOPs by 22.2%, while the mAP50 improves by 3.2%. Therefore, the proposed LAP-Net is validated as a reliable, competitive, and lightweight PCB defect detection model.