基于纹理特征的钢轨表面缺陷检测  

Rail Surface Defect Detection Based on Textural Features

作　　者：李刚 丁运峰 张宇豪 LI Gang;DING Yunfeng;ZHANG Yuhao(Anhui Vocational College of City Management,Hefei 230011,China;Wuhu GOOGOL Automation Technology Co.,Ltd.,Wuhu Anhui 241060,China;Nanjing Vocational Institute of Railway Technology,Nanjing 210031,China)

机构地区：[1]安徽城市管理职业学院,安徽合肥230011 [2]芜湖固高自动化技术有限公司,安徽芜湖241060 [3]南京铁道职业技术学院,江苏省南京市210031

出　　处：《兰州工业学院学报》2024年第4期70-76,共7页Journal of Lanzhou Institute of Technology

基　　金：芜湖市科技项目重点研发项目(022yf25)。

摘　　要：为提高钢轨表面缺陷识别的准确率,采用双边滤波的方法去除噪声,较好地保留缺陷边界;优化阙值计算方法,结合灰度直方图峰值的个数,采用相应的阙值计算方法,确保二值化结果可靠、有效;应用频域滤波优化,消除低频、细微的干扰区域;通过不变矩特征提取,有效地区分掉块和压溃缺陷,从而提高了钢轨表面缺陷识别的智能化程度和准确率。In order to improve the accuracy of rail surface defect identification,the method of bilateral filtering is used to remove noise and preserve the defect boundary.The threshold calculation method is optimized,combined with the number of gray histogram peaks,adopts the corresponding threshold calculation method to ensure that the binarization results are reliable and effective.Frequency domain filter optimization is applied to eliminate low frequency and subtle interference areas.Through the invariant moment feature extraction,the falling block and collapse defects can be distinguished effectively,which improves the intelligence level and accuracy of rail surface defect identification.

关 键 词：HALCON 不变矩 钢轨 表面缺陷 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]