\n \n \n \n \n 基于生成对抗网络数据增强与迁移学习的小样本激光软钎焊焊点缺陷检测研究 \n \n \t \n\t\n \n \n \n \n \n \n \n \n \n \n \n \n\t \n \n \n \n \n
\n
高级检索
\n
\n \n
\n
\n
\n
\n
\n \n \n
\n \n
\n \n
\n

基于生成对抗网络数据增强与迁移学习的小样本激光软钎焊焊点缺陷检测研究

Study on small-sample defect detection of laser soft soldering joints based on generative adversarial network data augmentation and transfer learning

\n \n \n \n \n \n
\n \n
\n \n \n
    \n \n \n
  • \n
    \n \n
    \n 摘要: \n 针对激光软钎焊焊点缺陷检测中样本稀缺、重叠敏感性高、目标微小及特征复杂等问题,提出一种融合生成对抗网络(generative adversarial network,GAN)数据增强与深度迁移学习的小样本检测方法.首先,基于集成视觉系统构建PCB(printed circuit board)焊点缺陷数据集;其次,设计结合混合注意力机制和谱归一化的辅助分类生成对抗网络(auxiliary classifier generative adversarial network,ACGAN)用于图像扩增,构建YOLOv5-TL迁移学习检测模型并优化迁移策略.试验表明,数据增强使非迁移模型精度提升约2.9%,迁移学习进一步提升约5.7%,两者结合提升约11.1%;数据增强略降收敛速度,迁移学习显著加快,结合后收敛速度提升约12.5%.该方法已成功部署于激光软钎焊设备,实现焊点缺陷实时在线检测,具备良好准确性与系统稳定性,满足实际在线缺陷识别需求. \n

     

    \n \n
    \n \n
    \n
    \n \n
    \n Abstract: \n To address the problems of sample scarcity, high overlapping sensitivity, small targets, and complex features in defect detection of laser soft soldering joints, a small-sample detection method integrating generative adversarial network (GAN) data augmentation and deep transfer learning is proposed. First, a printed circuit board (PCB) solder joint defect dataset was constructed based on an integrated visual system. Secondly, an auxiliary classifier generative adversarial network (ACGAN) combined with a hybrid attention mechanism and spectral normalization was designed for image augmentation, a YOLOv5-TL transfer learning detection model was constructed, and the transfer strategy was optimized. Experiments indicate that data augmentation improves the accuracy of the non-transfer model by approximately 2.9%, transfer learning further improves it by about 5.7%, and their combination achieves an improvement of around 11.1%; data augmentation slightly reduces convergence speed, transfer learning significantly accelerates it, and their combination improves convergence speed by about 12.5%. This method has been successfully deployed in laser soft soldering equipment to achieve real-time online defect detection of solder joints, and possesses good accuracy and system stability, meeting actual requirements for online defect recognition. \n

     

    \n \n
    \n \n
    \n
  • \n
    \n \n \n
  • \n
  • \n \n \n \n \n
  • \n \n
  • \n \n \n \n \n
  • \n
  • \n \n \n \n \n
  • \n
  • \n \n \n \n \n
  • \n \n \n \n \n \n
  • \n
  • \n \n \n
\n \n
\n
\n \n
\n \n \n
\n
\n \n \n \n \n\t \n \n\t \n\t
\n\t\t \n\t\t
\n\t\t\t
\n\t\t\t\t \n\t\t\t
\n\t\t\t \n\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

\n\t\t\t\t\t

\n\t\t\t\t
\n\t\t\t
\n\t\t\t \n\t\t\t

\n\t\t\t\t/ \n\t\t\t

\n\t\t\t \n\t\t\t
\n\t\t\t\t下载:  \n\t\t\t\t全尺寸图片 幻灯片 \n\t\t\t
\n\t\t
\n\t\t \n\t\t
\n\t\t\t
\n\t\t\t\t \n\t\t\t
\n\t\t\t
\n\t\t\t\t
    \n \n \n\t\t\t\t
\n\t\t\t
\n\t\t\t
\n\t\t\t\t \n\t\t\t
\n\t\t
\n\t\t \n\t\t
\n\t\t\t 返回文章 \n\t\t\t \n\t\t
\n\t
\n \n\t \n\t \n\t \n\t \n\t
\n\t\t
\n\t\t\t返回 \n\t\t
\n\t
\n\t \n\t
\n\t
\n\t\t\t \n\t
\n\t
\n\t\t\t \n\t\t\t\t \n\t\t\t \n\t
\n\t
    \n\t \n\t
\n\t
\n\t \n\t\t\t\t \n\t\t\t \n\t
\n\t
\n \n\n\n \n \n\t
\n\t\t
\n\t\t\t \n\t\t
\n\t\t \n\t
\n
\n \n \n \n \n