多种群联合的多目标进化自适应阈值图像分割算法  

Multi-population cooperation-based multi-objective evolutionary algorithm for adaptive thresholding image segmentation

作　　者：赵凤 郑月 刘汉强[2] 王俊 Zhao Feng;Zheng Yue;LiuHanqiang;Wang Jun(School of Telecommunication ＆ Information Engineering,Xi＇an University of Posts ＆ Telecommunications,Xi＇an 710061,China;School of Computer Science,Shaanxi Normal University,Xi＇ an 710119,China)

机构地区：[1]西安邮电大学通信与信息工程学院,西安710061 [2]陕西师范大学计算机科学学院,西安710119

出　　处：《计算机应用研究》2018年第6期1858-1862,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61571361;61102095;61202153;61671377);陕西省科技计划资助项目(2014KJXX-72);中央高校基本科研业务费专项资金资助项目(GK201503063);西安邮电大学西邮新星团队资助项目(xyt2016-1)

摘　　要：多阈值图像分割算法的阈值数目大多需要用户指定,人为干预较大。提出多种群联合的多目标进化自适应阈值图像分割算法,在多个分组种群的联合进化框架下,通过同时优化类间方差准则和模糊熵准则获得图像阈值,并在进化过程中采用自调节的交叉和变异操作产生子代种群并自动确定阈值数目。实验结果表明,该算法不仅能自适应得到合适的阈值数目,而且阈值分割效果也是比较理想的。The number of thresholds is mostly given in existing multi-level thresholding segmentation algorithms,which requires more human intervention. This paper proposed a multi-population cooperation-based multi-objective evolutionary algorithm for adaptive thresholding image segmentation. Under the framework of the multi-population cooperative evolution,this algorithm simultaneously optimized the between-cluster variance and fuzzy entropy to obtain the image thresholds,performed the self-tuning crossover and mutation operations on the chromosomes to generate the offspring population and automatically evolved the number of thresholds during the evolution process. Experimental results show that the proposed method not only can adaptively obtain the right number of thresholds,but also can achieve satisfactory thresholding segmentation results.

关 键 词：图像分割 多目标进化 多种群联合 自适应多阈值 自调节交叉和变异 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]