基于投影直方图的钢轨表面缺陷检测  

Detection of Rail Surface Defects Based on Projection Histogram

作　　者：陈思宇 张汝峰 常虹[2] 陈静 张松 Chen Siyu;Zhang Rufeng;Chang Hong;Chen Jing;Zhang Song(School of Electronic and Electrical Engineering,Cangzhou Jiaotong College,Cangzhou Hebei 061199;China Unicom Beijing Branch,Beijing 100089)

机构地区：[1]沧州交通学院电子与电气工程学院,河北沧州061199 [2]中国联通北京市分公司,北京100089

出　　处：《现代工业经济和信息化》2021年第11期19-20,76,共3页Modern Industrial Economy and Informationization

基　　金：2021年度河北省大中学生科技创新能力培育专项项目(基于机器视觉的铁路自动巡检方法研究);2020年沧州市科学计划自筹经费项目(204102004,204204002,204107009)。

摘　　要：针对钢轨表面离散缺陷检测算法复杂度高、精度低等问题,提出一种轻量级的钢轨表面缺陷检测算法。利用双边滤波和中值滤波对图像进行去噪,计算图像的投影直方图,并计算直方图的最值、均值和标准差,将直方图进行修正和线性拉伸,放大直方图变化的部分。根据直方图的均值和标准差选取合适阈值,检测出缺陷位置。实验表明:本算法对于重载钢轨缺陷检测正确率86.57%,普通钢轨缺陷检测正确率90.63%,因此本算法具有较高准确度,并具有一定工程性和应用性。Aiming at the high complexity and low accuracy of the detection algorithm for discrete defects on the rail surface,this paper proposes a lightweight rail surface defect detection algorithm.First use bilateral filtering and median filtering to denoise the image;then calculate the projection histogram of the image,and calculate the maximum value,mean and standard deviation of the histogram,correct and linearly stretch the histogram,and enlarge the histogram change Part;Finally,select an appropriate threshold according to the mean and standard deviation of the histogram to detect the defect location.Experiments show that the accuracy of this algorithm is 86.57%for heavy-duty rail defects and 90.63%for ordinary rail defects.Therefore,this algorithm has high accuracy,and has a certain degree of engineering and applicability.

关 键 词：钢轨缺陷 缺陷检测 投影直方图 

分 类 号：U216.3[交通运输工程—道路与铁道工程] TP391.4[自动化与计算机技术—计算机应用技术]