一种基于CNN的带钢表面缺陷识别方法  

Surface Defect Recognition of Strip Steel Based on CNN

作　　者：白贵龙 牛锐祥 Bai Guilong;Niu Ruixiang(Silicon Steel Business Unit of Shanxi Taigang Stainless Steel Co.,Ltd.,Taiyuan Shanxi 030002,China)

机构地区：[1]山西太钢不锈钢股份有限公司硅钢事业部,山西太原030002

出　　处：《山西冶金》2024年第6期32-34,共3页Shanxi Metallurgy

摘　　要：带钢表面质量是衡量产品性能的重要指标,准确识别带钢表面缺陷是带钢生产过程的关键一环,剪除缺陷带钢对于提升带钢成材率具有重要意义。为提升带钢表面缺陷识别的准确率,构建了基于CNN的带钢表面缺陷识别模型,通过多个卷积层提取图像特征,从而自动识别缺陷类别,实现了端到端的带钢表面缺陷识别过程。实验结果表明,CNN模型对于带钢表面缺陷识别准确率达到了96.5%,识别一张图片时间仅为1.5 ms,基本满足了带钢缺陷识别要求。The surface quality of strip steel is an important indicator for measuring product performance.Accurately identifying surface defects of strip steel is a key link in the production process of strip steel.Cutting off defective strip steel is of great significance for improving the yield of strip steel.In order to improve the accuracy of surface defect recognition for strip steel,a CNN based model for surface defect recognition of strip steel was constructed.Image features were extracted through multiple convolutional layers to automatically identify defect categories,achieving end-to-end surface defect recognition of strip steel.The experimental results show that the CNN model achieves an accuracy of 96.5%for identifying surface defects on strip steel,and the recognition time for one image is only 1.5 ms,which basically meets the requirements of strip steel defect recognition.

关 键 词：CNN 带钢表面 缺陷识别 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TG115[自动化与计算机技术—控制科学与工程]