基于多类支持向量机的X射线焊缝图像缺陷类型识别方法

压力容器的焊缝结构多变,纹路复杂,增加了压力容器检测焊缝缺陷时定级的难度.提出一种基于多类支持向量机的X射线焊缝图像缺陷类型识别方法.首先通过对X射线焊缝图像进行预处理及缺陷轮廓检测,提取表征焊缝缺陷的状态特征,以构造特征向量;然后基于多类支持向量机建立焊缝缺陷识别模型,对产品的焊缝缺陷进行分类识别.实验结果表明了该方法的有效性.

Weld Defects Identification Based on Multi-Class Support Vector Machine for X-Ray Inspection Digital Image

It is difficult to classify weld defects of pressure vessel because of complated welding structure and welding line. This paper presents a new weld defects identification scheme for X-ray inspection diginital image. Instead of using artificial neural network, the method reconginizes weld defects within X-ray image based on support vector machine(SVM) for pressure vessel. Above all, the edge of one weld defect is detected after X-ray weld image is preprocessed. Afterwards, eight defects features are extracted to construct one feature vector for each defect sample according to its edge figure. Furthermore, recongnizing model based on a multi-class SVM is established to classify weld defects by the feature vectors. The results showed the scheme could effectively identify weld defects genres for X-ray inspection diginital image.