Defects recognition on X-ray images for weld inspection using SVM
Xiao-Guang Zhang; Jian-Jian Xu; Guang-Ying Ge
Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on
Volume 6, Issue , 26-29 Aug. 2004 Page(s): 3721 - 3725 vol.6
Digital Object Identifier
Summary: Traditional manual assessment on weld defect has the shortcomings of troublesome operation and uncertain judgments. This paper presents a method of automatic defect recognition in weld image based on support vector machine (SVM). The method firstly preprocesses weld image and classifies the defects as 6 classes, then 8 features are selected according to the defect characteristics. Secondly, weld defects are classified by a multi-classifier based on SVM combined with the bintree. Finally, we compare the classifier based on multilevel SVM with the one based on fuzzy neural network (FNN) using total 84 samples in defect recognition. Experimental results show SVM has higher accuracy under the condition of small samples, and less effect on accuracy with the decrease of training samples. This research demonstrates that SVM has excellent performance for the defect recognition in weld image.
View citation and abstract |
Cart
