基于压缩感知的X射线螺旋焊管焊缝缺陷检测  

Detection for weld defects in spiral welded pipe by X-ray based on compressed sensing

作　　者：李勇[1] 高炜欣[1] 汤楠[1] 崔亚楠[1] 

机构地区：[1]西安石油大学陕西省钻机控制重点实验室,陕西西安710065

出　　处：《焊接技术》2013年第2期51-55,76,共5页Welding Technology

基　　金：陕西省自然科学基础研究计划项目(2010JQ8033)

摘　　要：应用压缩感知技术对X射线石油螺旋焊管焊缝缺陷进行识别。先对现场取得的焊管图像进行处理,截取出图像中的缺陷和噪声部分,在其中挑选出典型的缺陷和噪声作为研究样本。然后根据压缩感知技术训练出缺陷和噪声样本图像的冗余字典,通过该冗余字典求出待检测缺陷和噪声图像的稀疏表示。最后根据压缩感知图像识别原理进行待测图像的类型识别,得到待检测图像的缺陷和噪声识别率,也就是缺陷检测的灵敏度和特异度。检测结果表明,该方法更具有满足工程要求的高缺陷识别灵敏度和特异度。Weld defects detection in X-ray image of spiral welded pipe used in oil was identified through the compressed sensing technology.The field welded pipe images was processed firstly and weld defects and noises parts from those processed images was intercept.The typical defects and noises were selected as samples to construct a redundant dictionary of defects and noises separately.Owing to the redundant dictionary of defects and of noises,the sparse matrixes of defects under detection and noise images were calculated.Based on the image recognition principle using compressed sensing,above two sparse matrixes were used finally to acquire the recognition ratio of defects under detection and of noise image,which were also called defects detection sensitivity and specificity respectively.The detection results showed that this approach could get higher defect recognition sensitivity and specificity required in engineering applications.

关 键 词：压缩感知 缺陷识别 灵敏度 特异度 冗余字典 稀疏矩阵 

分 类 号：TG441.7[金属学及工艺—焊接]