基于厚壁工件X射线实时成像的焊缝缺陷自动检测

基于X射线成像的焊缝缺陷自动检测技术对提高工业射线检测的自动化水平具有重要意义。焊缝缺陷在线连续检测的实时性要求较高，随着工件厚度的增加，其焊缝X射线实时图像的信噪比变得很低，使得已有的处理算法难以满足实时性以及有效处理缺陷误检与漏检之间的矛盾。针对这些问题，在分析了传统方法在厚壁工件X射线图像焊缝缺陷自动检测中存在的问题基础上，对传统方法进行了改进，提出了双阈值背景消除法和平行焊接方向波形分析法，然后利用所提出算法之间的冗余性和互补性，融合多种分割结果以解决缺陷误检与漏检之间的矛盾。试验结果表明，所提出的缺陷自动检测方法能够在满足实时性要求的同时，实现缺陷检出，有效避免误检。

Automatic detection of weld defects based on X-ray images of thick-wall workpiece

The technology of weld defect automatic detection based on X-ray imaging plays an important role in improving the automation level of industrial radiography inspection. As the thickness of the weldment increases, the Signal-to-noise Ratio of the X-ray real-time image of the weldment is decreased. This reduces the timeliness of on-line continuous detection and makes the current method ineffective to deal with the con- flict between reducing false alarms and avoiding missed detections of weld defects. Based on the analysis of the drawbacks existing in traditional background subtraction and grey-level profile analysis method to detect weld defect in X-ray images of thick-wall weldment, the information fusion of multiple image segmentation algorithms is proposed to detect weld defects. Firstly, double threshold background subtraction and grey-lev- el analysis parallel to weld direction are proposed. Then, the segmentation results by different algorithms are fused to deal with the conflict of false alarms and missed detections. Experiment results show that the pro- posed method can meet the requirement of efficiency of on-line continuous detection of weld defects, and automatically detect weld defect of thick-wall weldment successfully.