基于聚类的埋弧焊X射线焊缝图像缺陷分割算法及缺陷模型

摘要

Regarding the present problems that the traditional image segmentation algorithm can only achieve a low successful defect segmentation ratio for the strong noise and low contrast of submerged-arc x-ray image,an efficient X-ray radiography image analysis algorithm is developed for the task of segmentation of submerged-arc welding defects.In the new algorithm,the defect is treated as noise and a new concept-＂gray density＂ is put forward for calculation convenience.Tested with 100 X-ray radiography images obtained from a real factory,the proposed algorithm can increase successful segmentation ratio and achieves a successful ration of 95%.Based on the clustering segmentation algorithm,a high dimension space defect mathematical model is presented.The model makes the characteristic of the complexity of the form into consideration.Real examples show that the model is effective and practical.The sensitivity curve of the presented clustering segmentation algorithm is also given.%针对埋弧焊X射线焊缝图像的噪声强、弱对比度特点和常规图像分割算法成功率低的现状,提出将缺陷视为噪声,利用密度聚类方法进行缺陷分割.在进行图像聚类时,提出图像灰度密度的概念,方便对焊缝图像的分割.通过对现场100张焊缝图像的试验表明,所提方法大幅度地提高了缺陷分割的成功率,将分割成功率提高至95%.在聚类分割算法基础上,通过试验给出一种新的高维空间缺陷数学模型,该模型综合考虑了缺陷形式复杂性等特征.通过试验在高维空间对模型予以验证,并结合所提聚类算法给出了覆盖率曲线.