基于X射线图像序列的焊缝缺陷自动检测方法

Automatic inspection of weld defects using X-ray image sequences

X射线无损检测对于焊接结构件的质量保证具有重要意义,焊缝缺陷X射线实时自动检测的核心技术是序列图像处理。目前,国内外的研究工作大多集中在基于孤立图像的处理中,由于焊缝缺陷形态的多样性而难以将噪声与微小缺陷区分开,因此无法解决缺陷的漏检率和误检率此消彼长的矛盾。该文着眼于时间序列图像,提出基于帧内分割和帧间跟踪相结合的焊缝缺陷检测方法,在图像的空间域处理中消除漏检,在图像的时间域处理中消除误检,从而为解决缺陷漏检和误检之间的矛盾提供了有效途径。试验和现场应用期间,对于常见的壁厚分别为8.8 mm、12.5 mm以5m/m in速度运动的螺旋焊缝钢管中的一系列自然缺陷,最小检出缺陷直径分别约为0.6mm、0.8mm,符合工业应用要求,与此同时无误检出现。

X-ray-based nondestructive testing （NDT） is used to assure weld quality. Image sequence processing is the key technology for automating weld defect detection. Most current research concentrates on the processing of a single static image. Automatic defect detection during continuous running welding is difficult due to misdetections and false detections since the methods were developed for single images. For image sequences, this paper presents a method for weld defect detection based on segmentation of individual images and tracking among the image sequence. The method reduces the false detection rate in time and the undetected ratio in the dimensional scale of the image processing to resolve the conflict between the misdetection ratio and the false detection ratio. Experiment and tests show that smallest detectable defects is 0. 6 mm in 8.8 mm thick running steel tubes and 0. 8 mm in 12. 5 mm thick tubes at 5 m/min with no false detections.