摘要
针对多阈值图像分割方法计算量大、分割精度低的问题,提出了基于改进正余弦算法(improved sine cosine algorithm,ISCA)的多阈值图像分割方法。首先对种群进行混沌初始化来提高初始种群质量;其次根据粒子适应度值的大小自适应地调整参数;最后引入反向学习策略并择优选取粒子。伯克利图像和植物冠层图像分割实验的结果表明,该算法的运行时间较短,而且分割精度较高,具有较强的鲁棒性。
Aiming at the problem of the computational complexity and low segmentation precision of the multi-threshold image segmentation method,this paper proposed an ISCA based multi-threshold image segmentation method.Firstly,this method used a chaotic initialization technique to improve the quality of initial population.Secondly,it introduced an adaptive strategy to adjust the parameters according to the fitness values.Finally,it utilized an opposition-based learning strategy,then selected the better particles.The results of the Berkeley image and the plant canopy image segmentation experiments show that this method has a satisfied performance in terms of running time and segmentation accuracy.And it has a strong robustness.
作者
郎春博
贾鹤鸣
邢致恺
彭晓旭
李金夺
康立飞
Lang Chunbo;Jia Heming;Xing Zhikai;Peng Xiaoxu;Li Jinduo;Kang Lifei(College of Mechanical&Electrical Engineering,Northeast Forestry University,Harbin 150040,China)
出处
《计算机应用研究》
CSCD
北大核心
2020年第4期1215-1220,共6页
Application Research of Computers
基金
中央高校基本科研业务费专项资金资助项目(2572019BF04)
国家自然科学基金资助项目(51609048)
黑龙江省研究生教育创新工程项目(JGXM_HLJ_2016014)。
关键词
正余弦算法
多阈值图像分割
混沌初始化
自适应
反向学习
sine cosine algorithm
multi-threshold image segmentation
chaos initialization
adaptive
opposition-based learning
