基于灰度共生矩阵的金属断口图像的分类研究  

Research of metal fracture image classification based on GLCM

作　　者：苏静[1] 黎明[1] 

机构地区：[1]南昌航空大学无损检测技术教育部重点实验室,南昌330063

出　　处：《计算机工程与应用》2008年第9期223-225,234,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60475002);无损检测技术教育部重点实验室开放基金

摘　　要：灰度共生矩阵在图像的纹理分析中是一个很重要的方法,但是其参数的选择现在还没有确定的方法。论文计算了灰度共生矩阵的13个特征参数,根据不同步长、不同灰度压缩级时特征值的变化曲线确定步长和灰度级别,然后根据相关系数分析选择其中的7个特征,利用加权欧式距离分类器将金属断口图像分为4类。与常用的5个特征进行分类相比,根据该方法选择的7个特征分类效果明显更好。GLCM is an important method in the texture analysis of image,but there isn＇t decided means to ascertain it＇s parameters until now.Thirteen parameters of GLCM are computed in this paper and then distance and gray level are selected according to feature＇s mutative curve of different distances and gray levels.At last,seven features are selected according to coefficient of correlation.Then weighted Euclidean distance classifier is designed to classify metal fracture image.Compared with five features in common use,the classification results using the seven features selected according to this method is much better.

关 键 词：金属断口 纹理分析 BP神经网络分类器 灰度共生矩阵 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]