基于图像特征的轨道表面缺陷识别研究
|
摘要
本文针对钢轨表面图像的特点,经过图像预处理、缺陷标记,提取出轨道表面存在的缺陷,计算得到缺陷的4个特征(矩形度、致密度、离心率和灰度均值),将获得的缺陷特征采用径向基神经网络进行分类,从而获得轨道表面缺陷的具体类别,进行自动报警供维修人员维修。仿真结果表明,基于图像的轨道缺陷识别能够精确的检测出轨道表面的疤痕、裂缝以及波纹擦伤缺陷。
Detecting the rail surface defects is one of the important means to guarantee safe transportation.According to the characteristics of the rail surface image,this paper uses advanced image processing technology to identify the rail surface defects.The surface defects are extracted after image preprocessing and defect marking,then 4 features of defect are calculated(rectangle degree, density,centrifugal rate and mean gray level),through which the specific defect categories are obtained by radial basis neural network for maintenance.Typical scars,cracks and corrugated abrasion defects are used to implement image defect feature extraction and classification in this paper.This simulation shows that the methods discussed can accurately detect the scars,cracks and corrugated abrasion defects of rail surface.
Detecting the rail surface defects is one of the important means to guarantee safe transportation.According to the characteristics of the rail surface image,this paper uses advanced image processing technology to identify the rail surface defects.The surface defects are extracted after image preprocessing and defect marking,then 4 features of defect are calculated(rectangle degree, density,centrifugal rate and mean gray level),through which the specific defect categories are obtained by radial basis neural network for maintenance.Typical scars,cracks and corrugated abrasion defects are used to implement image defect feature extraction and classification in this paper.This simulation shows that the methods discussed can accurately detect the scars,cracks and corrugated abrasion defects of rail surface.
引文
[1]孟繁顺.浅谈铁路钢轨断裂的原因及预防措施[J].铁道建筑技术,2011,(10):58-62.
[2]章华燕.钢轨擦伤检测算法研究[D].北京交通大学硕士学位论文,2016.
[3]Li Q,Ren S.A Real-Time Visual Inspection System for Discrete Surface Defects of Rail Heads[J].IEEE Transactions on Instrumentation and Measureme nt,2012,61(8):2189-2199.
[4]刘蕴辉,刘铁,王权良,罗四维.基于图像处理的铁轨表面缺陷检测算法[J].计算机工程,2007,(11):236-238.
[5]唐湘娜,王耀南.铁轨表面缺陷的视觉检测与识别算法[J].计算机工程,2013,(03):25-30.
[6]聂超.基于直方图的高效图像增强算法研究[D].杭州电子科技大学硕士学位论文,2014.
[7]李东兵,李国平,滕国伟,等.一种新的自适应维纳滤波方法[J].上海大学学报(自然科学版),2012,(06):555-560.
[8]时佳悦,张蕊萍,董海鹰,苟军年.基于灰度变化率的低对比度CT图像分割研究[J].兰州交通大学学报,2017,36(03):57-62[2017-09-04].
[9]付云凤.基于阈值的图像分割研究[D].重庆大学硕士学位论文,2013.
[10]沈夏炯,吴晓洋,韩道军.分水岭分割算法研究综述[J].计算机工程,2015,(10):26-30.
[11]江军兵,严利民.一种基于BP神经网络的磁钢表面缺陷分类方法[J].工业控制计算机,2016,(07):83-84+87.
[12]郭凌云.基于RBF神经网络的液体细分器的设计[D].北京邮电大学硕士学位论文,2013.
[2]章华燕.钢轨擦伤检测算法研究[D].北京交通大学硕士学位论文,2016.
[3]Li Q,Ren S.A Real-Time Visual Inspection System for Discrete Surface Defects of Rail Heads[J].IEEE Transactions on Instrumentation and Measureme nt,2012,61(8):2189-2199.
[4]刘蕴辉,刘铁,王权良,罗四维.基于图像处理的铁轨表面缺陷检测算法[J].计算机工程,2007,(11):236-238.
[5]唐湘娜,王耀南.铁轨表面缺陷的视觉检测与识别算法[J].计算机工程,2013,(03):25-30.
[6]聂超.基于直方图的高效图像增强算法研究[D].杭州电子科技大学硕士学位论文,2014.
[7]李东兵,李国平,滕国伟,等.一种新的自适应维纳滤波方法[J].上海大学学报(自然科学版),2012,(06):555-560.
[8]时佳悦,张蕊萍,董海鹰,苟军年.基于灰度变化率的低对比度CT图像分割研究[J].兰州交通大学学报,2017,36(03):57-62[2017-09-04].
[9]付云凤.基于阈值的图像分割研究[D].重庆大学硕士学位论文,2013.
[10]沈夏炯,吴晓洋,韩道军.分水岭分割算法研究综述[J].计算机工程,2015,(10):26-30.
[11]江军兵,严利民.一种基于BP神经网络的磁钢表面缺陷分类方法[J].工业控制计算机,2016,(07):83-84+87.
[12]郭凌云.基于RBF神经网络的液体细分器的设计[D].北京邮电大学硕士学位论文,2013.