学术论文

      冷轧带钢表面缺陷检测系统设计

      Design of Surface Defect Inspection System for Cold Rolled Strip

      摘要:
      针对现有带钢表面质量检测技术的缺陷检测精度与识别率不高的问题,以及冷轧带钢生产线的实际运行环境,设计了一套带钢表面缺陷实时检测系统,从系统整体设计、硬件结构、软件开发和检测分类算法等方面进行了深入的分析与研究.采用模块化软件设计理念、基于谱残差视觉显著性缺陷快速检测方法以及多分类器融合框架,实现了对带钢表面常见的孔洞、擦伤、氧化、边裂等不同类型和不同等级的缺陷进行精确、实时的检测与分类.实验结果表明:该算法缺陷检测率为94.68%,缺陷识别率达到了93.5%,单幅图像缺陷检测耗时仅为37.6 ms,符合当下高速运动带钢快速检测的需求.
      Abstract:
      As for the problems of the low precision and accuracy of the quality existing in the surface inspection techniques , an automated and real-time surface inspection system need to be developed for detection of the defects on the cold steel strip sur -face.Considering the operational aspects of the cold steel strip product line ,a deep research and analysis in the design of the whole system which contains structure of the hardware ,development of the software and the algorithms of the detection and classi-fication was proceeded .In this paper ,a fast detection method for the salient defects based on the spectral residual techniques was proposed.There are various types of the defects on the surface ,such as holes,scratches,oxide scales and silvers.Our method that combines the modular software design with the framework of multi-classifier fusion can detect and classify various types of defects with different hierarchy precisely in real time .Experiment show that our defect detection and recognition rate has reached 94.68%and 93.5%,respectively.Furthermore,it only takes 37.6 ms for a single image defect detection ,which meets the requirements of the fast detection for the high-speed rolling steel strips .
      作者: 徐森 [1] 屈尔庆 [1] 陈海永 [1] 刘坤 [2] 孙鹤旭 [1]
      Author: XU Sen [1] QU Er-qing [1] CHEN Hai-yong [2] LIU Kun SUN He-xu
      作者单位: 河北工业大学控制科学与工程学院,天津,300130 河北工业大学控制科学与工程学院,天津 300130;河北科技大学,河北石家庄 050000
      刊 名: 仪表技术与传感器 ISTICPKU
      年,卷(期): 2017, (1)
      分类号: TP391
      在线出版日期: 2017年3月31日
      基金项目: 国家自然科学基金项目,河北省自然科学基金项目,河北省高等学校科学技术研究项目,天津市特派员科技计划项目