用于焊缝检测的X射线实时图像序列时空域滤波  

Spatio-temporal filtering of real-time X-ray images for weld inspections

作　　者：侯润石[1] 都东[1] 邵家鑫[1] 朱新杰[1] 

机构地区：[1]清华大学机械工程系,先进成形制造教育部重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2009年第5期643-646,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(50875145)

摘　　要：X射线实时图像处理技术是在线自动检测焊缝缺陷的关键环节。由于原始图像噪声大,严重影响了焊缝缺陷X射线实时自动检测技术的实际应用。该文先对焊缝X射线实时图像序列进行匹配,在此基础上采用扩展的局部线性最小均方差法进行时空域滤波,从而综合利用X射线实时图像序列的时域信息和空域信息。实验结果表明:焊缝X射线实时图像时空域滤波效果明显优于现有的空域滤波和时域滤波,为后续焊缝缺陷检测的图像处理提供了良好条件。Real-time X-ray image processing is the key to on-line automatic inspection of weld defects. In practical applications, automatic weld inspections are greatly affected by big noise in the original image. This paper presents a method for analyzing real-time X-ray image sequences where the spatio-temporal filtering of the real-time weld images uses an extended local linear minimum mean square error method, so that the spatial and time domain information can be used together. Experimental results show that the denoising effect of the spatio-temporal filter is much better than conventional spatial domain filters and time domain filters. Thus, the image quality is greatly improved, which enhances subsequent image processing for weld defect detection.

关 键 词：X射线 焊缝缺陷 无损检测 图像匹配 时空域滤波 

分 类 号：TP274.51[自动化与计算机技术—检测技术与自动化装置]