一种基于模糊核聚类算法的图像分类方法  

An image classification method based on fuzzy kernel clustering algorithm

作　　者：王书文[1] 皮炳坤 张弘强[1] 马聪[1] 

机构地区：[1]西北民族大学电气工程学院,甘肃兰州730030

出　　处：《西北师范大学学报（自然科学版）》2016年第5期42-45,共4页Journal of Northwest Normal University(Natural Science)

基　　金：国家自然科学基金资助项目(61261042);西北民族大学中央高校基本科研业务费专项资金资助研究生项目(YXM2015215)

摘　　要：聚类分析是数据分析的一个重要方法.通过引用核函数,将核方法应用到模糊C均值(Fuzzy c-Means,FCM)算法中,优化FCM算法的目标函数,使样本点被非线性变换映射到高维特征空间进行聚类,不仅改善了聚类效果,而且增强了算法对噪声的鲁棒性.在真实样本集上进行了仿真实验,分类结果证实了该算法的有效性和普适性,因而是一种较为简单和实用的图像分类方法.The clustering analysis is a crucially important method for data analysis. By referring to kernel {unction, the kernel method is applied to the fuzzy C-means（fuzzy C-means, FCM） algorithm, optimizing objective {unction of FCM algorithm, sample points are mapped to high-dimensional feature space by the nonlinear transform cluster. The method can not only improve the clustering effect, but also enhance the algorithm robustness to noise. Simulated experiments are conducted on real sample set, the classification results prove the effectiveness and generalizability of the algorithm, so this algorithm is a relatively simple and practical method.

关 键 词：聚类分析 子空间聚类 核函数 核模糊聚类算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]