摘要
为了有效地使用用户给定的先验信息,并从多个角度考虑图像分割问题,提出了应用于彩色图像分割的半监督多目标进化模糊聚类算法。首先,将半监督方法引入到多目标进化聚类算法中,通过使用少量的监督信息指导聚类过程;其次,将最大熵正则化引入到带有监督信息的目标函数中,使目标函数具有清晰的物理意义;最后,利用监督信息构造基于相似性度量的有效指标从非支配解集中选择一个最优解。实验结果表明,该算法与传统的多目标进化聚类算法及半监督模糊聚类算法相比具有更好的灵活性和准确性。
In order to more effectively using prior information associated with data obtained satisfactory results,this paper proposed semi-supervision multi-objective fuzzy evolutionary clustering application in color image segmentation.First,this paper introduced a semi-supervised approach as a multi-objective evolutionary clustering algorithm,utilized a small amount of given prior knowledge to guide the clustering process.Second,it introduced maximum entropy as a regularized term in its objective functions with monitoring information such that its resulting formulas had the clear physical meaning.Lastly,it produced a set of non-dominated solutions in the final generation,from which the best solution in terms of a proposed validity index BI based on similarity measure was chosen to be the best clustering solution.The experimental results show that feasibility and accuracy of the proposed method compared with semi-supervised fuzzy algorithms and traditional multi-objective fuzzy evolutionary clustering have better effective.
作者
杨颖青
赵凤
Liu Wanjun;Li Tianhui;Qu Haicheng;Dong Shuaihan;Yin Xiu;Yang Yingqing;Zhao Feng(College of Software,Liaoning Technical University,Huludao Liaoning 125105,China)
出处
《计算机应用研究》
CSCD
北大核心
2018年第10期3126-3129,3163,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(61571361
61102095)
陕西省科技计划资助项目(2014KJXX-72)
关键词
彩色图像分割
半监督
多目标进化算法
最大熵
color image segmentation
semi-supervision
multi-objective evolutionary algorithm
maximum entropy
