期刊文献+

基于改进YOLOv5的焊缝缺陷检测算法

Weld Defect Detection Algorithm Based on Improved YOLOv5
下载PDF
导出
摘要 焊接作为工业生产的重要一环,其优劣关乎最终产品质量。为了解决传统人工目视方法准确度与效率欠佳的情况,提出一种基于YOLOv5的焊缝缺陷检测算法YOLOv5-Z。YOLOv5-Z算法的改进可以分为以下几点:首先,焊缝缺陷种类丰富,为了精准检测微小焊缝,在基准网络中加入微小瑕疵检测层;其次,为了加强网络对于特征的利用效率,改善提取特征质量,在网络中Neck部位插入多头注意力;最后,构建一套工业场景焊缝及缺陷数据集,以完成焊缝缺陷的训练与测试。经过实验验证,所提YOLOv5-Z网络平均精度mAP达到98.15%,满足实际工业场景焊缝缺陷检测的需求。 As an important part of industrial production,welding is related to the quality of the final product.In order to solve the problem of poor accuracy and efficiency of traditional manual visual method,a weld defect detection algorithm YOLOV5-Z based on YOLOv5 was proposed.The improvement of YOLOv5-Z algorithm can be divided into the following points:First of all,there are many types of weld defects.In order to accurately detect micro welds,micro defect detection layer is added to the benchmark network.Secondly,in order to enhance the efficiency of feature utilization and improve the quality of feature extraction,multihead attention is inserted in the Neck part of the network.Finally,a set of weld and defect data set in industrial scenes is constructed to complete the training and testing of weld defects.Through experimental verification,the average accuracy mAP of the proposed YOLOv5-Z network reaches 98.15%,which meets the requirements of weld defect detection in actual industrial scenes.
作者 张路 ZHANG Lu(School of Electromechanical Engineering,Dalian Minzu University,Dalian 116605,China)
出处 《山东工业技术》 2023年第6期97-103,共7页 Journal of Shandong Industrial Technology
关键词 焊缝缺陷 YOLOv5 小目标 多头注意力 weld defect YOLOv5 small target multi-head attention
  • 相关文献

参考文献2

二级参考文献10

共引文献149

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部