摘 要:模糊C- 均值(Fuzzy C-Means,FCM)聚类算法是一种基于划分的无监督聚类算法,也是较为常见的图像分割算法之一,该算法通过寻找0 ~ 1 之间的模糊隶属度等级来进行图像分割,并通过在特征空间中寻找聚类中心来达到最小化目标函数的目的。它的局限性主要有实时性较差、初始聚类中心的设置对最终结果影响较大、未考虑空间因素导致抗噪性弱。本文将mini-batch 方法应用到FCM 算法中,加快了FCM 算法的收敛速度,提高了算法的效率及时效性,一定程度上解决了当数据特征复杂、集合较大时,FCM 算法的实时性不是很理想的问题,继而节省算法运行的时间。
关键词:FCM 聚类;mini-batch;图像分割
中图分类号:TP391.41 文献标识码:A 文章编号:2096-4706(2019)19-0015-03
The Mini-batch FCM Algorithm for HD Color Image Segmentation
NI Cui1,LI Qian2,XUAN Jiahui1
(1.Jiangsu Automation Research Institute,Lianyungang 222061,China;2.School of Information Engineering,Lianyungang Technical College,Lianyungang 222006,China)
Abstract:Fuzzy C-Means (FCM) clustering algorithm is an unsupervised clustering algorithm based on partition. It is also one of the common image segmentation algorithms. This algorithm conducts image segmentation by looking for the fuzzy membership grade between 0 ~ 1. The objective function is minimized by finding the clustering center in the feature space. Its limitations mainly include poor real-time performance,large impact on the final results due to the setting of the initial clustering center,and weak noise resistance due to the absence of space factors. In this paper,the mini-batch method is applied to the FCM algorithm to accelerate the convergence speed of the FCM algorithm,improve the efficiency and timeliness of the algorithm,and to some extent solve the problem that the real-time performance of the FCM algorithm is not ideal when the feature set of data is large,and then save the algorithm time.
Keywords:FCM clustering;mini-batch;image segmentation
参考文献:
[1] LLOYD S. Least squares quantization in PCM [J].IEEE Transactions on Information Theory,1982,28(2):129-137.
[2] MACQUEEN J. Some methods for classification and analysis of multivariate observations [C]//Proc. of 5th Berkeley Symposium on Mathematical Statistics and Probability,USA:University of California Press,1967:281-297.
[3] DUNN J C. Well-Separated Clusters and Optimal Fuzzy Partitions [J].Journal of Cybernetics,2008,4(1):95-104.
[4] BEZDEK J C. A convergence theorem for the fuzzy ISODATA clustering algorithm [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1980,2(1):1-8.
[5] BEZDEK J C,EHRLICH R,FULL W. FCM:The fuzzy c-means clustering algorithm [J].Computers & Geosciences,1984,10(2-3):191-203.
[6] DEKEL O,GILAD-BACHRACH R,SHAMIR O,et al.Optimal Distributed Online Prediction using Mini-Batches [J].2012,13:165-202.
作者简介:
倪翠(1992-),女,汉族,宁夏中卫人,助理工程师,硕士研究生,研究方向:机器学习中的优化方法;
李千(1970-),男,汉族,江苏灌云人,副教授,硕士,研究方向:网络大数据分析、嵌入式系统应用;
玄甲辉(1987-),男,汉族,山东泰安人,工程师,硕士研究生,研究方向:智能装备系统与电子信息系统。