埋弧焊X射线焊缝图像缺陷分割检测技术  

Real-time detection and segmentation of submerged-arc welding defects in X-ray radiography images

作　　者：高炜欣[1] 胡玉衡[2] 穆向阳[1] 武晓萌[1] 

机构地区：[1]西安石油大学电子工程学院 [2]威斯康星大学麦迪逊分校电气与计算机工程系

出　　处：《仪器仪表学报》2011年第6期1215-1224,共10页Chinese Journal of Scientific Instrument

基　　金：陕西省自然科学基础研究计划(2010JQ8033);陕西省教育厅专项科研计划(08jk411);西安市科技计划(YF07033)资助项目

摘　　要：针对在埋弧焊X射线焊缝图像的强噪声和弱对比度特点下常规图像分割算法成功率低的现状,通过综合的分析和实验有针对性地给出了系统、实用的缺陷分割方法。首先通过实验给出了一种提取焊缝图像感兴趣区域(the region of interest,ROI)的方法。该方法通过中值滤波,基于sin函数的图像增强、大津法分割、Sobel算子边缘检测和Hough变换可以定量计算出X射线焊缝图像的ROI区域。进一步通过实验给出基于大津法的焊缝缺陷分割算法,实验表明在无人工设定初始分割阈值情况下这一分割算法具有较高的分割成功率。为进一步提高分割成功率,针对焊缝缺陷相对面积较小的特点,提出将缺陷视为噪声,将平均局部平均灰度视为密度,利用密度聚类方法进行缺陷分割。该方法在78张有缺陷的焊缝图像中,成功地分割出74张图像中的缺陷。最后在所研究算法基础上给出了一个综合考虑到各种因素的分割算法,即可节约计算时间,又可以保证分割的成功率。An efficient X-ray radiography image analysis algorithm is developed for the real time detection and segmentation of submerged-arc welding defects.The raw X-ray image is first pre-processed to remove impulsive noise,enhance gray scale contrast with sin function,and segment the region of interests（ROI） where the welding image locates.Next,a segmentation method based on OTSU is studied,and examples show that the method can obtain good segmentation result.Lastly,a density based spatial clustering method is developed for defect detection and segmentation.Compared with a baseline threshold segmentation algorithm,the proposed clustering segmentation method demonstrates significant performance enhancement.100 X-ray radiography images obtained from a real factory were tested;and the proposed algorithm achieves a sensitivity of 74/78 and specificity of 19/21.Compared with previously reported algorithms,the proposed algorithm is robust,efficient and directly applicable to real world applications.

关 键 词：焊接 缺陷 分割 图像处理 聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]