改进YOLOv5s的钢板表面缺陷检测算法  

Steel-Plate Surface-Defect Detection Algorithm Based on Improved YOLOv5s

作　　者：周彦[1] 孟江南 吴佳[1] 罗智 王冬丽[1] Zhou Yan;Meng Jiangnan;Wu Jia;Luo Zhi;Wang Dongli(School of Automation and Electronic Information,Xiangtan University,Xiangtan 411105,Hunan,China;Hunan Valin Xiangtan Iron and Steel Co.,Ltd.,Xiangtan 411105,Hunan,China)

机构地区：[1]湘潭大学自动化与电子信息学院,湖南湘潭411105 [2]湖南华菱湘潭钢铁有限公司,湖南湘潭411105

出　　处：《激光与光电子学进展》2023年第4期383-391,共9页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(61773330);湖南省国家应用数学中心项目(2020YFA0712503);湖南省教育厅科研项目(19C1740);湖南省科技计划(2020GK2036);上海市科委项目(19511120900)。

摘　　要：针对传统方式检测钢板表面缺陷存在检测精度低、检测速度慢的问题,提出一种改进YOLOv5s算法。首先,使用基于交并比(IoU)度量距离的K-means算法对钢铁数据集进行重新聚类,获得多组锚框,通过遗传算法对其进行变异运算,得到与全体标注框更匹配的多组锚框;其次,在Mosaic数据增强上融合MixUp,抑制过拟合,提升模型的泛化能力;然后,对网络结构进行改进,融入注意力模块,进一步提高了网络的特征提取能力;最后,针对难识别样本,在损失函数中融入Focal loss,提高网络的收敛速度与检测精度。实验结果表明,改进后的YOLOv5s算法在测试集上的平均精度均值(mAP)可达78.4%,比原始的YOLOv5s算法提高了3.0个百分点,速度上与原始YOLOv5s基本持平。所提算法在保持高检测速度的基础上,检测性能也优于DDN、Faster R-CNN和YOLOv3。To solve the problem of the low accuracy and slow speed of traditional methods for detecting surface defects in steel plates,we propose an improved YOLOv5s algorithm.First,the steel datasets were reclustered using Kmeans algorithm based on the intersectionoverunion(IoU)metric distance,to obtain multiple groups of anchor boxes;a genetic algorithm was used to perform mutation operations and obtain multiple groups of anchor boxes that match the entire ground truth box better.Second,MixUp was fused with the Mosaic data enhancement to avoid overfitting and improve the generalizability of the model.Then,the network structure was improved,and an attention module was incorporated to improve feature extraction capability of the network further.Finally,Focal loss was incorporated into the loss function to improve the convergence speed and detection accuracy of the network for hardtoidentify samples.Our experimental results show that the mean average precision(mAP)of the improved YOLOv5s algorithm on a test set is 78.4%,which is 3.0 percentage points higher than that of the original algorithm,and the speed is same as the original YOLOv5s.The detection performance of the improved YOLOv5s algorithm is better than that of DDN,Faster RCNN,and YOLOv3,and it maintains a high detection speed.

关 键 词：YOLOv5s 钢板表面缺陷检测 注意力机制 Focal loss 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]