基于显著性模型的磁瓦表面缺陷检测  

Surface defect detection of magnetic tile based on significance model

作　　者：陈森 谌伦超 秦大辉[2] Chen Sen;Chen Lunchao;Qin Dahui(Sichuan Third Surveying&Mapping Engineering Institute,Chengdu 610599,China;School of Civil Engineering&Geomatics,SouthWest Petroleum University,Chengdu 610500,China)

机构地区：[1]四川省第三测绘工程院,成都610599 [2]西南石油大学土木工程与测绘学院,成都610500

出　　处：《黑龙江科技大学学报》2024年第5期700-708,共9页Journal of Heilongjiang University of Science And Technology

基　　金：岩土力学与工程国家重点实验室开放基金项目(Z020026);四川省青年科技创新科研团队项目(2019JDTD0017)。

摘　　要：针对磁瓦图像由于光照不均匀、背景纹理复杂等造成的缺陷类型难以检测的问题,提出了一种基于改进的同态滤波和MT显著性融合模型的磁瓦表面缺陷检测方法,将改进的同态滤波函数分离图像高频、低频分量,与采用限制对比度直方图均衡法增强的高频和低频图像线性叠加,采用MT显著性融合的通用模型,检测不同类型的磁瓦缺陷。结果表明,利用形态学闭运算和噪点去除算法得到最终的检测缺陷,文中方法正确检测率达92.95%。该方法可更加准确地检测不同类型的磁瓦表面缺陷。This paper proposes a magnetic tile surface defect detection method based on improved homomorphic filtering and the MT(Magnetic tile) saliency fusion model to address the hard deeetection of defect types of the magnetic tile image due to uneven illumination and complex background texture.The study involves using an improved homomorphic filtering function to separate the high-frequency and low-frequency components of the image,linearly overlaying the high-frequency and low-frequency images enhanced by the limited contrast histogram equalization method,using a universal model of MT significance fusion to detect different types of the magnetic tile defects.The result shows that the final defect detection is obtained by using morphological closure and noise removal algorithms,the proposed method can effectively and conrrectly detect the different types of magnetic tile surface defects,with a correct detection rate of 92.95%.

关 键 词：表面缺陷检测 显著性检测 同态滤波 AC模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]