列车车轮踏面缺陷在线磁粉检测

Online magnetic particle inspection of defect on train wheel tread

荧光磁粉探伤技术应用于列车车轮表面与近表面出厂缺陷检测,当前主要依靠人工识别缺陷,提出了一种将机器视觉应用于磁粉探伤,实现列车车轮踏面自动化缺陷识别的方法。首先通过分离三原色通道的灰度化预处理,消除紫光反射噪声;使用中值滤波与形态学开运算,抑制图像噪声并加强形态;采用多阈值最大类间方差分割算法解决大背景下阈值偏移问题;提取灰度特征和形状特征以及基于旋转不变等价局部二值模式的纹理特征,使用随机森林构建分类模型。缺陷分类准确率为97.63%,单幅图片处理时间69.95 ms,实验结果表明,该方法能有效快速识别列车车轮踏面缺陷,满足列车踏面在线缺陷检测的自动化要求。

Fluorescent magnetic particle inspection technology is applied to the defect detection on the surface and near-surface of train wheels and it mainly relies on manual identification at present.This paper proposes a method of applying machine vision to magnetic particle inspection to realize automatic defect recognition of train wheel treads.Separating the color channels firstly to eliminate the purple light reflection noise;Based on median filtering and morphological opening operation,the image is denoiseed and the shapes are intensified;Using the Maximum Inter-Class Variance Multi-Threshold Algorithm to solve the problem of threshold shift because of the large background;Texture features based on the equivalent rotation invariant Local Binary Patterns,gray features as well as shape features are extracted and the image is classified by the Random Forest Method.The defect classification accuracy is 97.63%,and the single frame detection takes only 69.95 millisecond,Experimental results show that this method can effectively and quickly identify train wheel tread defects,which meets the automation requirements of online detection on train wheel tread.