钢吊车梁疲劳裂缝开展的图像识别方法研究  

Research on image recognition method for fatigue crack development in steel crane girders

作　　者：朱文俊[1] 黄竞崛 Zhu Wenjun;Huang Jingjue(Baoshan Iron&Steel Co.,Ltd.,Shanghai 201900,China;Baosteel Zhanjiang Iron&Steel Co.,Ltd.,Zhanjiang Guangdong 524072,China)

机构地区：[1]宝山钢铁股份有限公司,上海201900 [2]宝钢湛江钢铁有限公司,广东湛江524072

出　　处：《山西建筑》2024年第14期48-52,59,共6页Shanxi Architecture

摘　　要：钢结构服役过程中,结构受到反复的疲劳荷载,即使未达到极限强度,结构仍会产生表面疲劳裂缝,发生破坏。目前钢结构裂缝检测技术已有很多研究,包括人工巡检、接触式智能涂层传感技术等。但人工巡检效率低下,且极易漏检,接触式的技术操作复杂、成本较高。近年来,利用图像识别技术检测结构缺陷的方法由于其易操作性和较高的检测效率,吸引了缺陷检测研究者们的关注。深度学习方法能够很好地解决复杂环境中的特征匹配问题。因此,采用深度学习的方法检测钢吊车梁疲劳裂缝,训练基于YOLOv5网络的裂缝图像识别模型,并通过传统图像处理与深度学习的结合方法,提高检测的精度,最终设计并实现了基于PC端与Android端的钢结构板件裂缝检测软件的开发制作。During the service of steel structure,the structure is subjected to repeated fatigue loading,and even if the ultimate strength is not reached,the structure wi still produce surface fatigue cracks and damage.At present,there have been many researches on the crack detection technology of steel structure,including manual inspection,contact intelligent coating sensing technology and so on.However,manual inspection is inefficient and very easy to miss detection,and the contact type technology is complicated to operate and high cost.In recent years,the method of detecting structural defects using image recognition technology has attracted the attention of defect detection researchers due to its ease of operation and high detection efficiency.Deep learning methods can well solve the feature matching problem in complex environments.Therefore,a deep learning approach is used to detect fatigue cracks in steel crane beams,train a crack image recognition model based on YOLOv5 network,and improve the accuracy of detection by combining the traditional image processing and deep learning methods,and ultimately design and implement a crack detection software for steel structural panels based on PC and Android.

关 键 词：钢结构 疲劳裂缝 深度学习 图像识别 

分 类 号：TU375.1[建筑科学—结构工程]