摘要
图像分割是一类需要在非线性参数空间中寻求最优解的有约束非线性优化问题.为提高此类优化问题的寻优精度,提出了一种基于鸽群优化算法的图像分割方法.首先以分割阈值为优化变量,将图像分割建模为以最大间类方差为优化目标,以像素概率分布有限为约束条件的非线性优化问题;随后,以随机的分割阈值作为迭代初值,采用鸽群优化算法(PIO)求解最优参数;最后,利用所得最优解作为最佳阈值实现图像分割.为验证方法的有效性,分别对具有两类不同特征的图片进行分割实验,并采用重叠度及时间效率对算法进行评估,进一步与PSO、KSW智能优化算法对比.结果表明,该算法重叠度最高,运算时间最短.并且对算法中的参数进行修改,将图像分割结果进一步优化.
Image segmentation was a kind of constrained nonlinear optimization problem that aimed to seek the optimal solution in nonlinear parameter space.In order to improve the precision of the optimization problem,an image segmentation method based on pigeon group optimization algorithm was proposed.Firstly the segmentation threshold was used as the optimization variable,and the image segmentation was modeled as a nonlinear optimization problem with the optimal threshold equation as the objective function,and the inter-class variance and the w0 and w1 ranged as the constraints.Then,using random segmentation threshold as the initial value of iteration,the optimal parameters were solved by the pigeon group optimization algorithm(PIO).In order to verify the validity of this method,two kinds of images with different features were divided into experiments,and the algorithm was evaluated by overlapping degree and time efficiency.The results showed that the algorithm had the highest degree of overlap and the shortest operation time.
作者
胡春鹤
王依帆
朱书豪
刘文定
HU Chunhe;WANG Yifan;ZHU Shuhao;LIU Wending(School of Engineering,Beijing Forestry University,Beijing 100083,China)
出处
《郑州大学学报(工学版)》
CAS
北大核心
2019年第4期42-47,共6页
Journal of Zhengzhou University(Engineering Science)
基金
中央高校基本科研业务费专项资金资助项目(2016ZCQ08)
北京林业大学2018教育教学研究项目(BJF02018JYZD004)
关键词
鸽群优化
图像分割
群体智能优化
阈值分割
图像处理
pigeon group optimization
image segmentation
group intelligence optimization
threshold segmentation
image processing