一种冷轧带钢表面缺陷检测模型  

A Surface Defect Detection Model for Cold Rolled Strip Steel

作　　者：曾艳 吴泽启 张天有 李佳瑶 石惠文 ZENG Yan;WU Zeqi;ZHANG Tianyou;LI Jiayao;SHI Huiwen(Tangshan Polytechnic University,Tangshan 063299,China;Research and Development Center of Intelligent Manufacturing and Maintenance Application Technology for EMUs in Universities of Hebei Province;Hebei University of Science and Technology,Baoding 071000,China)

机构地区：[1]唐山工业职业技术大学,河北唐山063299 [2]河北省高校动车组智能制造与运维应用技术研发中心 [3]河北科技学院,河北保定071000

出　　处：《工业技术与职业教育》2025年第2期13-18,共6页Industrial Technology and Vocational Education

基　　金：教育部高等学校科学研究发展中心项目“激光切割钢板智能码垛机器人研究”(课题编号:2023DT020),主持人吴泽启。

摘　　要：带钢作为一种重要的原材料已经应用到各个行业,其质量的优劣直接关系到最终产品的性能与质量。为了更准确地检测带钢表面缺陷,控制带钢质量,提出了一种冷轧带钢表面缺陷的检测模型。该模型在YOLOv5框架下进行改进,主要有3个方面:通过引入注意力机制模块以增强特征抽取架构;通过采用SIOU(Sum of Intersection over Union)损失函数来优化模型训练过程;通过改进置信度预测的损失函数以增强模型在识别真实对象时的精准度。经过实验验证,所提出的改进模型可以有效地进行冷轧带钢表面缺陷的检测,且与同类算法YOLOv4和YOLOv5相比,检测的平均准确度都有所提升。As an important raw material,strip steel has been applied in various industries,and its quality directly affects the performance and quality of the final product.In order to detect the surface defects of strip steel effectively,and control the strip steel surface quality,a detection model is proposed in this paper.The model has been improved under the YOLOv5 framework,which has three main improvements:(1)Enhancing the feature extraction architecture by introducing an attention mechanism module;(2)Optimizing the model training process by using the SIOU(Sum of Intersection over Union)loss function;(3)Improving the loss function of confidence prediction to enhance the accuracy of the model in identifying real objects.The experimental results show that the proposed model can detect the surface defects in cold-rolled strip steel effectively,and the average detection accuracy of the proposed model has been improved,comparing with the similar algorithms YOLOv4 and YOLOv5.

关 键 词：带钢表面缺陷 YOLOv5 注意力机制 SIOU损失函数 置信度预测损失函数 

分 类 号：TN911.73[电子电信—通信与信息系统]