SVMs在油气管道焊缝缺陷检测中的应用  

Application of SVMs in the detection of oil/gas pipeline weld defects

作　　者：武晓朦[1] 高炜欣[1] 袁磊[2] 刘畅 

机构地区：[1]西安石油大学陕西省钻机控制重点实验室,陕西西安710065 [2]巴陵石化环氧树脂事业部电气车间,湖南岳阳414014 [3]西安航空动力股份有限公司无损检测中心,陕西西安710021

出　　处：《西安石油大学学报（自然科学版）》2011年第6期97-101,12,共5页Journal of Xi’an Shiyou University（Natural Science Edition）

基　　金：陕西省自然科学基础研究计划(编号:2010JQ8033);陕西省教育厅专项科研计划项目(编号:09JK699);西安市科技计划项目(编号:YF07033);中石油科技创新基金研究项目(编号:2010D-5006-0601)

摘　　要：提出了基于支持向量机的焊管焊缝缺陷识别方法和步骤.对焊缝图像采用了图像增强、形态学处理、边缘检测等操作,解决焊缝图像在输入时受到外界干扰带来的图片噪声过多、缺陷边缘模糊和对比度低等问题,便于对焊缝图像进行特征提取.结合焊缝缺陷样本多分类的特点,对焊缝图像进行分类时使用SVM"一对一"聚类结构并对样本进行识别.实验结果表明,该模型具有识别精度高、速度快、容易实现等优点,适合对焊管焊缝缺陷的识别.A new method based on support vector machines(SVMs) for the identification of oil/gas pipeline weld defects was proposed.A series of processing to the weld image actions was carried,such as image enhancement,morphological processing and edge detection.As a result,the problems of excessive noise,fuzzy edge and low contrast can be solved in order to enhance the quality of the weld image and be easy to extract the characteristics of the weld image.Because of the many types of weld defects,the one-to-one classification structure of SVMs was adopted to identify the defects of sample.The experimental results show that the proposed model has the advantages of high identification speed and accuracy and easy implementation.And it is very suitable for the identification of the pipeline weld defects.

关 键 词：油气管道 焊缝缺陷 缺陷检测 图像分割 特征提取 模式识别 支持向量机 

分 类 号：TE973.6[石油与天然气工程—石油机械设备]