基于改进YOLOv5算法的带钢表面缺陷检测  

Steel Strip Surface Defect Detection Based on Improved YOLOv5 Algorithm

作　　者：程豪 蒋占四[1] 陈晓鑫 郑洪鑫 蒋慧 CHENG Hao;JIANG Zhansi;CHEN Xiaoxin;ZHENG Hongxin;JIANG Hui(School of Mechanical and Electrical Engineering,Guilin University of Electronic Technology,Guilin 541004,China;School of Civil and Engineering Management,Guangzhou Maritime University,Guangzhou 510725,China)

机构地区：[1]桂林电子科技大学机电工程学院,桂林541004 [2]广州航海学院土木与工程管理学院,广州510725

出　　处：《组合机床与自动化加工技术》2023年第10期141-144,149,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(51565008);桂林电子科技大学研究生教育创新计划项目(2022YCXS009)。

摘　　要：针对当前带钢在表面缺陷检测过程中存在检测算法精度有待提高等问题,提出了一种基于改进YOLOv5算法的带钢表面缺陷检测模型。首先,在检测端构建新的检测层,提高网络对不同尺寸目标的检测;其次,在主干网络结构中引入注意力模块,进一步加强网络提取特征的能力;然后,通过BiFPN_Add来增强深浅层特征信息的融合;最后,构建新的CNeB模块来取代各检测层对应的C3模块,进而增强网络对特征的提取。实验结果表明,改进后的算法在NEU-DET数据集上均值平均精度达到了80.9%,较原有的算法提升了4.5%,同时检测速度与原模型保持基本不变,性能优于目前其他主流的检测方法。Aiming at the problem that the accuracy of the detection algorithm needs to be improved in the process of surface defect detection of steel strips,a surface defect detection model of steel strips based on improved YOLOv5 algorithm is proposed.Firstly,a new detection layer was constructed at the detection end to improve the detection of objects of different sizes.Secondly,the attention module was introduced into the backbone network structure to further strengthen the ability of the network to extract features.Then,BiFPN_Add was used to enhance the fusion of deep and shallow feature information.Finally,a new CNeB module is constructed to replace the C3 module corresponding to each detection layer,so as to enhance the feature extraction of the network.Experimental results show that the mean average precision of the improved algorithm on NEU-DET dataset reaches 80.9%,which is 4.5%higher than that of the original algorithm.At the same time,the detection speed remains basically the same as that of the original model,and the performance is better than other mainstream detection methods.

关 键 词：缺陷检测 YOLOv5 注意力机制 特征融合 

分 类 号：TH162[机械工程—机械制造及自动化] TG65[金属学及工艺—金属切削加工及机床]