炭素制品缺陷X射线自动检测关键技术的研究  

Study of Key X-ray Automatic Inspection Techniques for Carbon Product Defects

作　　者：周贤[1] 刘义伦[1] 赵先琼[1] 

机构地区：[1]中南大学机电工程学院,长沙410075

出　　处：《机械科学与技术》2006年第11期1358-1361,共4页Mechanical Science and Technology for Aerospace Engineering

基　　金：湖南省教育厅重点科研项目(03A052);企业横向项目(G1999064910)资助

摘　　要：以炭素制品X射线检测图像为对象,对其关键技术进行了研究。设计了目标边界提取算法和基于小波变换的图像增强算法等预处理方法。采用数学形态学和迭代阈值分割相结合的方法进行缺陷提取,取得了良好的效果。制定了一套用于特征描述的参数,设计了基于遗传算法的特征选择策略,实现了对缺陷原始特征量的优化选择。利用BP神经网络分类器及选择的特征值对缺陷进行了模式分类。实验结果表明,本文提出的方法比较有效,可以用于实际缺陷的自动提取与识别。The paper studied key techniques for the X-ray inspection image of a carbon product. The target boundary extraction algorithm, the image enhancement algorithm based on wavelet transformation and other preprocessing methods were designed, Mathematical morphology in combination with iteration threshold segmentation was used to extract defects. A set of parameters for feature description was prepared; feature selection strategies based on genetic algorithm were designed; the original feature numbers of defects were selected optimally; their pattern classification was carried out through using BP neural network classification devices and the features selected. An experiment indicates that the techniques have good effects on the automatic recognition and extraction of actual defects.

关 键 词：炭素制品 X射线图像 缺陷提取 自动识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]