摘要
模糊C-均值聚类(FCM)算法由于能够很好地解决像素分类的不确定性而得到广泛应用,但是聚类中心的初始化对其分割效果有很大的影响。文中以初始聚类中心为重点,分别用K均值算法、遗传算法、蚁群算法、粒子群优化算法优化FCM算法初始聚类中心,将优化后的结果作为FCM的初始聚类中心,并利用MATLBA软件进行了实验仿真。通过实验结果对比分析,不仅优化后的运算时间有所减少,而且所得的聚类中心更加稳定,使得分割出的目标也更加完整、清晰,验证了改进算法的有效性。
Fuzzy C-Means clustering （ FCM） algorithm was widely used because of effectively solving the uncertainty of pixel classifica-tion,but the initialization of clustering center had a great influence on its segmentation effect. So,focused on initial clustering center,re-spectively use K-means algorithm,genetic algorithm,ant colony algorithm,particle swarm optimization algorithm to optimize the FCM algorithm initial clustering center and take optimized result as its initial clustering center. MATLBA software was applied to carry out the experimental simulation. Through comparing and analyzing the results of the experiment,not only reduce the optimized operation time, but the clustering center is more stable,making segmentation effect is more complete and clear,which verifies the validity of the improved algorithm.
出处
《计算机技术与发展》
2015年第5期17-20,共4页
Computer Technology and Development
基金
甘肃省自然科学基金项目(1310RJZA057)
