直缝钢管X射线图像焊缝缺陷检测降噪算法  

X-ray image noise reduction in the defect detection of straight-weld steel-pipe

作　　者：石涵[1] 都东[1] 邹怡蓉[1] 邵家鑫[1] 王力[1] 

机构地区：[1]清华大学机械工程系,北京100084

出　　处：《中国科技论文》2013年第8期717-720,共4页China Sciencepaper

基　　金：高等学校博士学科点专项科研基金资助项目(20090002110080)

摘　　要：X射线实时成像技术广泛应用于直缝钢管焊缝缺陷射线检测系统。与螺旋焊缝相比,直缝钢管的厚度大、X射线图像噪声大。针对直缝厚壁钢管X射线图像高噪声、低对比度的特点,研究了直缝钢管X射线图像的噪声特点,拟合了噪声方差与图像灰度值的线性关系曲线,利用噪声方差与图像灰度值的关系估计图像局部噪声方差,并以此提出针对直缝厚壁钢管X射线图像的自适应邻域滤波算法。实验结果表明,改进的滤波算法明显优于中值滤波和原始的自适应邻域滤波算法,它在改善图像质量和降低图像噪声的同时,能较好地保存焊缝边缘和缺陷边缘等图像重要细节,为后续直缝钢管X射线图像焊缝缺陷自动检测提供了理论基础。X-ray imaging technology is widely used in defect detection of straight weld. Compared with spiral weld, straight weld has the characteristics of higher noise and lower image contrast, which goes against the detection of weld defects. In this paper, the characteristics of noise in straight weld X-ray images is investigated. Based on the shot noise analysis, an adaptive filter for straight weld is proposed to filter the images. Results show that the proposed straight weld noise reduction algorithm is much better than the mean filter and the median filter; it cannot only reduces the image noise, but also save the important image details such as the edge of the weld and defects. It also helps the manual defect detection and provides a theoretical basis for the develop- ment of automatic detection system.

关 键 词：焊缝 缺陷 无损探伤 图像降噪 

分 类 号：TG441.7[金属学及工艺—焊接]