改进人工蜂群优化的K均值图像分割算法
2018-09-05李海洋何红洲
李海洋 何红洲



文章编号: 2095-2163(2018)03-0045-05中图分类号: 文献标志码: A
摘要: 关键词: (School of Information Engineering, Mianyang Normal University, Mianyang Sichuan 621000, China)
Abstract: In the image segmentation, K-means clustering algorithm has the disadvantage of low accuracy and poor stability. The hybrid algorithm based on artificial bee colony and K-mean tend to be too inefficient to meet the application requirements. To cure the above problems, a new image segmentation algorithm called IABC-K is proposed in this study. The artificial bee colony algorithm is improved according to its different characteristics in honey source renewal and mining. An adaptive neighborhood search mechanism associated with optimal fitness is adopted to improve the speed of honey source renewal. A linear decreasing neighborhood search strategy associated with optimal fitness is adopted to improve the quality of honey source mining. Experimental results show that IABC-K algorithm is superior to other similar algorithms in terms of quality, efficiency and stability. IABC-K algorithm has better image segmentation quality and higher image segmentation efficiency. It can be applied in image processing field with high quality and performance requirements.
Key words:
基金項目:
作者简介:
收稿日期: 引言
K均值聚类算法是聚类问题研究的经典算法,常应用于图像分割[1]。但K均值聚类算法的聚类结果往往依赖于初始值的选取,图像分割质量较低[2]。针对K均值聚类算法的缺点,很多研究人员采用进化技术对K均值聚类算法进行优化。进化技术中,主要包括有:遗传算法[3]、粒子群优化[4-5]、人工蜂群优化[6](artificial bee colony, ABC)等。其中,ABC算法性能优异,全局寻优能力强,能比遗传算法、粒子群优化算法更快收敛于最优解[7]。
在ABC算法优化K均值聚类的研究中,梁冰等[8]将与当前维度最优解差值的变化率作为权值,对蜜源的搜索公式进行了改进,提高了算法的鲁棒性和聚类精度。宋锦等[9]修改了ABC算法对蜜源的更新方法,获得了较好的图像分割质量。Shokouhifar等[10]利用ABC算法来调整模糊规则,使算法具有自适应特性。Cong等[11]通过变异操作增强了ABC算法的搜索能力,提高了算法的聚类质量。Bose等[12] 使用一种模糊隶属函数来搜索ABC算法最优解,用以初始化K均值聚类中心,使算法在收敛性、时间复杂度、鲁棒性和图像分割精度方面表现较好。赵文昌等[13]利用距离最大最小乘积方法对ABC种群进行初始化,并采用自适应搜索参数调整邻域搜索范围,得到了较好的图像分割效果。但是,上述混合聚类算法大都针对特定条件,并存在稳定性欠佳的问题,不能达到聚类质量和聚类效率同时提高的目的,因而难以应用在分割质量和分割效率要求较高的图像处理领域。……
