基于双向加权特征融合网络的铸件内部缺陷检测方法  

Casting Internal Defect Detection Method Based on Bidirectional Weighted Feature Fusion Network

作　　者：王蕾[1,2] 贺万山 张泽琳 夏绪辉 WANG Lei;HE Wan-shan;ZHANG Ze-lin;XIA Xu-hui(Key Laboratory of Metallurgical Equipment and Control,Ministry of Education,Wuhan University of Science and Technology,Wuhan430081,Hubei,China;Hubei Provincial Key Laboratory of Mechanical Transmission and Manufacturing Engineering,Wuhan University of Science and Technology,Wuhan430081,Hubei,China)

机构地区：[1]武汉科技大学冶金装备及其控制教育部重点实验室,湖北武汉430081 [2]武汉科技大学机械传动与制造工程湖北省重点实验室,湖北武汉430081

出　　处：《铸造》2024年第6期843-851,共9页Foundry

基　　金：国家自然科学基金面上项目(52275503);湖北省重点研发计划项目(2022BAD102,2023BAB048)。

摘　　要：针对X射线无损探伤过程中铸件内部缺陷小、对比度弱、人工识别效率低等问题,提出了一种基于双向加权特征融合网络的铸件内部缺陷检测方法。在YOLOv5网络模型基础上引入改进的坐标注意力模块(NCA),以提高网络对不规则缺陷和小缺陷的学习能力;引入双向特征金字塔网络(BiFPN)代替原有路径聚合网络(PANet),以实现缺陷特征多尺度高效融合,并使用EIoU Loss回归损失函数提高缺陷边界框定位的精度。试验结果表明,本文所提方法对铸件内部小目标、弱对比度缺陷具有良好的检测性能。Aiming at the problems of small internal defects,weak contrast and low efficiency of manual recognition in the process of X-ray nondestructive testing,a method of casting internal defects detection based on biweighted feature fusion network was proposed.Based on the YOLOv5 network model,an improved coordinate attention module(NCA)was introduced to improve the learning ability of the network for irregular defects and minor defects.Bidirectional feature pyramid network(BiFPN)was introduced to replace the original path aggregation network(PANet)to achieve multi-scale efficient fusion of defect features,and EIoU Loss regression loss function was used to improve the accuracy of defect boundary frame location.The experimental results showed that the proposed method had good performance in detecting small targets and weak contrast defects in the castings.

关 键 词：铸件 缺陷检测 深度学习 注意力模块 双向加权特征融合 

分 类 号：TG115[金属学及工艺—物理冶金]