基于改进YOLOv8s的热轧带钢表面缺陷检测方法  

Hot Rolled Strip Surface Based on Improved YOLOv8s Defect Detection Method

作　　者：朱小慧 李国伟[1] 苌道方[1] 杜吉旺 ZHU Xiaohui;LI Guowei;CHANG Daofang;DU Jiwang(School of Logistics Engineering,Shanghai Maritime University,Shanghai 201306,China;Hudong-Zhonghua Shipbuilding(Group)Co.,Ltd.,Shanghai 200129,China)

机构地区：[1]上海海事大学物流工程学院,上海201306 [2]沪东中华造船(集团)有限公司,上海200129

出　　处：《船舶工程》2025年第1期124-131,共8页Ship Engineering

摘　　要：[目的]为解决传统表面缺陷检测方法存在的速度慢、效率低、终端设备计算资源有限等问题,[方法]对基于改进YOLOv8s的热轧带钢表面缺陷检测方法进行分析。[结果]研究表明:相较于原始YOLOv8s模型,改进的YOLOv8s模型的检测精度与模型轻量化程度都得到了显著提升;相较于目前主流的表面缺陷检测算法,改进的YOLOv8s模型在热轧带钢表面缺陷检测中可有效提高小目标检测能力,其轻量化模型有利于在移动端或嵌入式设备上进行部署。[结论]研究成果可为热轧带钢表面缺陷检测提供一定参考。[Purpose]To address the problems of slow speed,low efficiency,and limited computational resources of traditional surface defect detection methods,[Method]the surface defect detection method for hot-rolled strip steel based on improved YOLOv8s is analyzed.[Result]Research shows that compared with the original YOLOv8s model,the improved YOLOv8s model significantly improves the detection accuracy and model lightweighting;Compared with the current mainstream surface defect detection algorithms,the improved YOLOv8s model can effectively improve the small target detection capability in the surface defect detection of hot-rolled steel strip.Its lightweight model is suitable for use in mobile or embedded devices.[Conclusion]The research results can provide some references for the surface defect detection of hot-rolled strip steel.

关 键 词：YOLOv8s 热轧带钢 表面缺陷检测 轻量化 动态卷积 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]