埋弧焊X射线焊缝缺陷图像分类算法研究

对埋弧焊X射线焊缝圆形和线形缺陷图像进行分析,针对焊缝缺陷局部图像强噪声、弱对比度和常规方法不易区分类型的特点,将主成成分分析的思想引入焊缝圆形和线形缺陷类型分类。分析缺陷疑似局部图像自相关矩阵特征值发现,圆形线形焊缝缺陷疑似局部图像分类问题可降维为一维问题,极大地简化计算和提高运算速度。基于此给出圆形和线性缺陷分类算法,由于将缺陷图像分类问题降维,使得分类算法对模板的选择具有较强的鲁棒性。通过现场超过400张焊缝缺陷局部图像的实验表明,无论如何选取模板,线性缺陷的识别率均在98%以上,圆形缺陷的识别率在89%~98.8%之间,且在16次模板更换实验中,4次圆形缺陷识别率达到98.8%。

Sub arc X ray welding defect image classifying algorithm

The characteristics of circular defect and linear defect of sub arc x ray images are analyzed. In order to overcome shortcomings of low contrast and big noise of the radiographs. The idea of principal component analysis is introduced into the algorithm for classifying the kind of circular and linear defects. Eigenvalues of auto covariance matrix for suspected defect regions are analyzed, which show that the suspected defect region classifying problem can be reduced to a one dimension classifying problem. The dimension reduction in classifying can simplify the calculation procedure and improve the calculation speed. The algorithm for classifying circular and linear defect is also given. Because the classifying problem is reduced to a one dimension problem, the choice of model images has comparatively little influence on classifying results. No matter how to select model images, the real example shows that the recognition ratio for linear defect reaches 98%~99%, and the recognition ratio for circular defect reaches 89%~98.8%.