| 赵欣洋， 蔡超鹏， 王思， 刘志远.基于深度学习的不规则特征识别检测技术[J].轻工机械,2019,37(3):60-65 |
| 基于深度学习的不规则特征识别检测技术 |
| Irregular Feature Recognition and Detection Technology Based on Deep Learning |
| |
| DOI：10.3969/j.issn.1005 2895.2019.03.012 |
| 中文关键词: 金属轴 不规则缺陷 无损检测 深度学习 |
| 英文关键词:nondestructive testing irregular defects metal axes deep learning |
| 基金项目: |
|
| 摘要点击次数: 1741 |
| 全文下载次数: 2108 |
| 中文摘要: |
| 〖HK44〗〖HT5”H〗摘要：〖HT5”K〗针对目前工业上金属轴零件在加工的过程中可能由于加工失误、本身材质等原因产生不同缺陷，而传统的检测方法检测精
度和泛化能力有限的现状，课题组提出了基于深度学习的不规则特征识别技术，来提升对金属轴表面缺陷的检测效率。课题组设计了金属轴表面缺陷图像预处理
方法，提升采集的缺陷图像的质量；对传统深度学习Faster R CNN进行改进，设计了模型的特征提取网络、RPN网络、分类网络以及模型参数，提升模型的检测
性能。实验结果表明本技术能有效提升工业流水线对金属轴缺陷的检测效率和精度，可同时检测多种不同种类的缺陷。课题组的研究成果具备良好的泛化能力。 |
| 英文摘要: |
| In the process of processing and production，irregular defects may occur due to processing errors，material quality and other reasons，however，
the traditional defect detection methods for metal axle detection have some shortcomings，such as low recognition accuracy，lack of generalization ability and so on. The non destructive
testing method based on deep learning in this paper was put forward，it could improve the detection efficiency of metal shaft surface defects.In order to improve the quality of defect image
， the research team designed the image preprocessing method of surface defect of metal shaft，improved the traditional deep learning Faster R CNN，designed the feature extraction
network，RPN network，classification network and model parameters of the model，and improved the detection performance of the model. The experimental results show that this
technology can effectively improve the detection efficiency and accuracy of metal axis defects in industrial pipeline，it can detect various kinds of defects. This research has good
generalization ability. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
