基于DR的封头电子束焊缝缺陷自动识别方法研究

Research on automatic identification method of dome electron beam weld defects based on DR

在现有的DR焊缝缺陷自动识别方法中,一般都是通过大数据分析缺陷特征得出缺陷模型,而此类方法在样本数目不够的情况下误判率很高。为解决此类问题,提出了一种DR图像本体比较自动识别新方法。首先对检测原始图像进行预处理使图像满足评片要求。其次截取焊缝区域图像进行处理,放大缺陷特征,再将图像灰度化处理并滤波,利用三西格玛准则对缺陷位置进行判定,并可根据缺陷位置像素点得出缺陷面积。检测结果表明焊缝区域DR图像经图像处理及滤波后能放大缺陷特征,可以实现对封头电子束焊缝X射线DR图像的缺陷自动识别,最小可发现直径在0.4mm的气孔及宽度在0.2mm的裂纹缺陷,为此类电子束焊DR检测提供了图像缺陷自动识别的解决方案。

In the existing methods of automatic defect identification for DR weld,defect models are usually derived from large data analysis of defect characteristics,and this method has a high misclassification rate when the number of samples is insufficient.To solve this problem,a new method for automatic recognition of DR image ontology comparison is presented.First,the original image is detected and preprocessed to make the image meet the evaluation requirements.Secondly,the image of the welding seam area is intercepted and grayscale processed,and the grayscale data is differentiated,the defect features are enlarged,and the true and suspected defects are extracted using the three Sigma criterion.Finally,the location of the defect is extracted by the extreme difference method to eliminate the interference of the suspected defect.The results show that the combination of grey-scale signal difference method and extreme difference method can be used to automatically identify the defects in the X-ray DR image of the electron beam welding seam on the head.At the minimum,the pores with a diameter of 0.4 mm and crack defects with a width of 0.2 mm can be found,which provides a solution for automatic image defect recognition for this kind of electron beam welding DR detection.