基于支持向量机的射线检测焊接图像中缺陷识别  

Defects Recognition Based on Support Vector Machine within Radiographic Testing Weld

作　　者：刘元祥[1] 张晓光[1] 高顶[1] 

机构地区：[1]中国矿业大学机电工程学院

出　　处：《煤矿机械》2006年第5期773-776,共4页Coal Mine Machinery

基　　金：江苏省博士后科研基金资助课题(2004035);中国矿业大学科技基金资助课题(2005B005)

摘　　要：提出了应用支持向量机(SVM)进行射线检测焊接缺陷识别的方法。该方法首先对图像进行预处理,并根据缺陷特点提取、选择8个参数作为特征参数,将焊缝内常见缺陷分为6类,根据有限的学习样本,建立影响缺陷类别的条件、因素和类别之间的一种非线性映射,对测试的样本进行识别。According to the problem of defect recognition within X - ray inspection weld, the method using SVM to recognize weld defects is put forword. This method prepmcesses images firstly, and extracts and chooses 8 parameters as feature parameters according to the characteristics of defects, which are divided into 6 classes. Using the limited learning samples, a non - linear map is built up between the defect classes and the conditions and factors affecting them, and then testing samples can be recognized based on this map.

关 键 词：支持向量机 多类分类 焊接缺陷 识别 射线检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TG441.7[自动化与计算机技术—计算机科学与技术]