太赫兹科学与电子信息学报2025，Vol.23Issue(7)：720-730,735,12.DOI:10.11805/TKYDA2023362
基于改进YOLOv5的复合材料夹芯结构缺陷的太赫兹检测与识别
Terahertz detection and recognition of defects in composite sandwich structures based on improved YOLOv5
摘要
Abstract
During the preparation process of composite sandwich structures,various defects may occur,which can lead to safety hazards during use.Terahertz non-destructive testing technology has been proven to be an effective means for detecting defects in composite sandwich structures.However,in the defect detection process,due to issues such as the blurred edges and low resolution of terahertz images,there are problems of missed and false detections of defects.To improve the detection accuracy and achieve automatic defect detection,an improved YOLOv5 network model is proposed to detect defects.The specific improvements include adding Convolutional Block Attention Mechanism(CBAM)at different positions in the Backbone module to enhance the network's anti-interference ability and obtain more defect detail information;adding two cross-scale connection paths to the Neck end according to Bi-directional Feature Pyramid Network(BiFPN)to increase the information transfer between different network layers,and replacing the C3 module at the Neck end with the improved GhostC3 module to reduce parameters and computational load while maintaining accuracy,enabling rapid and accurate detection of defects.The results show that the mean Average Precision(mAP)of the improved YOLOv5 is 90.4%,which is 0.9%higher than that of the original network.Meanwhile,the number of parameters and Floating-point Operations(FLOPs)are reduced by 18.25%and 12.66%respectively,accelerating the model training process and enabling more precise and rapid detection of defects.关键词
无损检测/复合材料/太赫兹时域光谱系统/YOLOv5网络/缺陷检测Key words
Non-Destructive Testing(NDT)/multi-layer structure/Terahertz Time-Domain Spectroscopy(THz-TDS)system/YOLOv5/defect detection分类
通用工业技术引用本文复制引用
王淑杰,李向东,朱东胜,徐丽,柴玫,杨秀蔚..基于改进YOLOv5的复合材料夹芯结构缺陷的太赫兹检测与识别[J].太赫兹科学与电子信息学报,2025,23(7):720-730,735,12.基金项目
山东省科技型中小企业创新能力提升工程资助项目(2022TSGC2178) （2022TSGC2178）
淄博市重点研发计划资助项目(2021XCCG0067) （2021XCCG0067）
