摘要
为了提高X射线焊缝缺陷分类的识别率,提出将拉普拉斯特征映射(Laplacian Eigenmaps,LE)算法与支持向量机(Support Vector Machine,SVM)算法相结合应用于焊管焊缝缺陷检测。该方法首先提取焊缝缺陷的形状和几何特征,然后通过极大似然估计值(Maximum Likelihood Estimation,MLE)法对提取的7维样本特征进行本征维数估计并利用拉普拉斯特征映射法对各类特征向量进行降维处理,最后分别采用支持向量机和BP神经网络进行分类对比实验。实验表明,基于SVM及LE降维的X射线焊缝缺陷分类算法具有较高的识别精度,平均分类准确率达93%,能够较准确地对焊缝缺陷进行分类识别。
In order to improve the recognition rate of X-ray weld defect classification,it is proposed to apply Laplacian Eigenmaps（ LE） algorithm and support vector machine（ SVM） algorithm to the weld seam defect detection. Firstly,the shape and geometrical features of the weld defects are extracted. Then the feature dimensions of the extracted 7-dimensional samples are estimated by the maximum likelihood estimation method,and the dimension of the feature vectors is reduced by using Laplacian feature mapping method. Finally,the weld defects are classified using support vector machine and BP neural network separately and the classification results are compared. The experimental results show that the X-ray weld defect classification algorithm based on SVM and LE has higher recognition precision,and its average classification accuracy is 93%. The weld defects can be accurately classified using this method.
作者
王雒瑶
高炜欣
王欣
WANG Luoyao GAO Weixin WANG Xin(Key Laboratory of Shaanxi Province for Gas and Oil Well Logging Technology,Xi＇an Shiyou University,Xian 710065 ,Shaanxi ,China Key Laboratory of MOE for Photoelectrical Oil and Gas Logging and Detecting, Xi＇an Shiyou University,Xian 710065, Shaanxi, China)
出处
《西安石油大学学报（自然科学版）》
CAS
北大核心
2017年第5期96-101,106,共7页
Journal of Xi’an Shiyou University（Natural Science Edition）
基金
2013陕西省自然科学资助项目(2013JQ8049)
陕西省教育厅重点实验室科研计划项目(14JS079)
2013陕西省教育厅自然科学专项(2013JK1077)
陕西省自然科学基础研究计划青年人才项目(2015JQ5129)
陕西工业攻关计划项目(2013K09-18)
西安石油大学研究生创新与实践能力培养项目(YCS16212056)
