改进蜂群算法的图像阈值分割方法  

Image threshold segmentation method based on improved artificial bee colony

作　　者：张海涛[1,2] 程新文 熊红伟[1,3] 马海荣 陈联君[1] 钱小刚[1,4] 

机构地区：[1]中国地质大学信息工程学院,武汉430074 [2]河南财经政法大学资源与环境学院,郑州450034 [3]中冶集团武汉勘察研究院有限公司,武汉430080 [4]漳洲市测绘设计研究院,福建漳州363000

出　　处：《计算机应用研究》2017年第12期3880-3884,共5页Application Research of Computers

基　　金：高分辨率对地观测重大专项资助项目(07-Y30A05-9001-12/13)

摘　　要：为快速高效地进行图像分割,针对人工蜂群算法存在的收敛速度慢、易陷入局部最优解等问题,提出了一种基于改进人工蜂群算法分割二维Otsu图像的新方法。通过对蜜源更新过程中向当前最优蜜源方向进行引导,可以加快算法的收敛速度;为避免算法陷入局部最优并加快收敛速度,在局部搜索过程中逐步缩减了搜索范围并加入了放弃机制;针对较大梯度值无意义的问题,限定了蜜源范围,以提高算法的效率。最后结合具有不同直方图分布的图像进行了实验,结果表明了算法稳健、高效、快速的特性。In order to segment images exactly and quickly and for the problems of poor at convergence speed and easy fall to local best in artificial bee colony algorithm( ABC),this paper proposed a new method based on an improved ABC algorithm segmenting two dimensional Otsu images. In the nectar update procedure,it guided the search direction to the current best nectar to speed up the convergence speed of the algorithm. In order to avoid the algorithm falling into a local best solution and accelerate the convergence speed,it reduced the search range gradually and implemented the abandoning mechanism in the local search procedure. Considering meaningless problems of a large gradient vaule,it limited the nectar range in nectar initialization and updating process. At last,experimental results on images with different histogram distribution show that the algorithm characteristics is robust,efficient and fast.

关 键 词：图像分割 人工蜂群算法 最大类间方差法 局部搜索 二维直方图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]