中值改进的分割算法在钢缺陷检测中的应用  

Application of Image Segmentation Based on Class Median in Steel Defect Detection

作　　者：张文识[1] 孙杰[1] 张长军[1] 程新求 

机构地区：[1]辽宁石油化工大学机械工程学院,辽宁抚顺113001 [2]中石油昆仑天然气利用有限公司苏里格分公司,内蒙古鄂尔多斯017300

出　　处：《当代化工》2013年第9期1331-1333,共3页Contemporary Chemical Industry

摘　　要：针对压力钢管内壁表面缺陷图像在直方图呈偏态或重尾时图像分割效果差的问题,在传统的图像分割思想的基础上,提出了一种基于灰度中值和灰度平均离差分生成图像最佳分割阈值主变量的改进算法。实验结果表明,改进后的算法对于能够稳定有效地分割压力钢管内壁正常部位与缺陷部位,并将缺陷部位的细节特征增强显示,为后续分析缺陷原因提供了很好的形态学依据;显示了其在化工设备的完整性检测技术中的应用意义。In allusion to the poor results of segmentation in steel defects detection when the histogram of images of defects was skew and heavy-tailed,an improved method for image segmentation based on class median was proposed according to the conventional segmentation algorithms： minimum error thresholding that employed the sample mean value,the convergence process of optimization was also discussed based on particle swarm optimization.The testing results indicate the modified method can highlight the details of defects’ morphology,which can provide sound grounds for cause analysis;meanwhile,the improved algorithm is applicable in the detection of the integrity of the chemical equipments.

关 键 词：管壁缺陷检测 图像分割 灰度中值 最小误差法 

分 类 号：TG356[金属学及工艺—金属压力加工]