Scholarlink

<p>A machine vision system for detecting surface defects on engine blocks in an engine manufacturing line is proposed in this paper. This system utilizes deep learning algorithms to address the issues of high defect omission rate, high false alarm rate, and slow detection speed in the current inspection process. Specifically, an improved pix2pix generative adversarial networks model is proposed by introducing a ResNet network to improve the feature expression capability and generate images with lower Frechet inception distance score. Furthermore, the generated images are used to augment the training dataset and combined with the YOLOv5 convolutional neural network to enhance the training effect on small sample data. This method is compared with Faster-RCNN and single shot detector detection networks to validate its superiority in defect recognition performance. Experiments show that the proposed system achieves higher accuracy and robustness, with an average recognition accuracy of 98.4% and a detection time less than 0.5&#xa0;s per image. The system has been applied in the actual production line and achieved good performance, demonstrating its feasibility and potential for practical use in the manufacturing industry.</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

On-line surface defects detection of engine blocks based on improved CNN + GAN

  • 期刊论文

  • 发表日期：

    2025年2月17日

查看全文
  • Zhilin Zeng,
  • Hao Qu,
  • Jun Lv,
  • Zhengchun Du

摘要

A machine vision system for detecting surface defects on engine blocks in an engine manufacturing line is proposed in this paper. This system utilizes deep learning algorithms to address the issues of high defect omission rate, high false alarm rate, and slow detection speed in the current inspection process. Specifically, an improved pix2pix generative adversarial networks model is proposed by introducing a ResNet network to improve the feature expression capability and generate images with lower Frechet inception distance score. Furthermore, the generated images are used to augment the training dataset and combined with the YOLOv5 convolutional neural network to enhance the training effect on small sample data. This method is compared with Faster-RCNN and single shot detector detection networks to validate its superiority in defect recognition performance. Experiments show that the proposed system achieves higher accuracy and robustness, with an average recognition accuracy of 98.4% and a detection time less than 0.5 s per image. The system has been applied in the actual production line and achieved good performance, demonstrating its feasibility and potential for practical use in the manufacturing industry.