基于Gabor变换和分形特征的铁谱图像聚类分割  

Clustering Segmentation of Ferrography Based on Gabor Transform and Fractal Feature

作　　者：郭恒光[1] 瞿军[2] 

机构地区：[1]海军航空工程学院研究生管理大队,山东烟台264001 [2]海军航空工程学院飞行器工程系,山东烟台264001

出　　处：《小型微型计算机系统》2014年第6期1392-1395,共4页Journal of Chinese Computer Systems

摘　　要：针对铁谱图像的分割问题,为了充分利用其分形特征,提出结合Gabor变换和分形特征的铁谱图像聚类分割算法.设计一组Gabor滤波器,对铁谱图像进行滤波,得到多个尺度、多个方向的滤波后的图像,计算每幅图像的分形特征,并结合图像的灰度特征,构成铁谱图像的特征向量.为了减少特征之间的信息冗余,降低算法运行时间,采用核主成分分析法对特征向量进行降维.然后利用模糊C均值聚类方法实现图像分割.最后的算例表明提出的算法能够充分利用磨粒、磨粒链及背景区域分形特征的差别,快速、准确地达到铁谱图像分割的目的.Aiming at the segmentation problem of ferrography, and making full use of ferrography fractal feature, clustering segmenta- tion algorithm that based on Gabor transform and fractal feature was put forward. After being filtered by Gabor filter bank, multi-scale and multi-orientation filtered images were got. Fractal feature of every image was calculated pulsing grey feature, ferrography feature vector was acquired. In order to decrease the dimension of features greatly and reduce the complexity of computation, the kernel principal component analysis method was used. Then the fuzzy C-means clustering method was employed to achieve ferrography segmentation. At last, example shows that the proposed algorithm can make full use of fractal feature differences between wear particle, wear particle chain and background area, and can segment ferrography quickly and accurately.

关 键 词：铁谱图像分割 GABOR变换 分形特征 模糊C均值聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]