摘要
为了解决传统模糊C均值聚类算法(FCM)在对复杂图像进行分割时目标函数收敛速度较慢和算法优化结果依赖于算法初始聚类中心选择的问题,本文提出了采用蜻蜓算法优化FCM迭代过程的策略,优化后的算法能够更迅速更稳定地收敛于全局最优解。为验证优化算法的实用性与正确性并探究优化的实质,从伯克利标准图像库随机选择10幅图像进行实验,用其他元启发式算法优化的FCM算法与原FCM算法作为对照,比较分割的精确度与时耗,并记录各个部分代码的耗时进行比较。实验结果表明:在FCM算法迭代的过程中,时间开销最大的部分是计算图像的隶属度矩阵。蜻蜓算法在面对复杂问题时能减少隶属度矩阵的计算次数,更快搜索到全局最优解,从而达到优化。算法通过牺牲一部分精度,大幅降低耗时,适用于需要快速处理复杂图像的工作。
In order to solve the problems that the target function has a slow convergence rate when using traditional fuzzy C-means clustering(FCM)algorithm to segment complex images and that the result of algorithm optimization depends on the initial clustering center selection of the algorithm,this paper presents a strategy to optimize the FCM iteration process using Dragonfly algorithm.The optimized algorithm can converge to the global optimal solution more quickly and stably.To verify the practicability and correctness of the optimization algorithm and to explore the essence of the optimization,10 images are randomly selected from the Berkeley Standard Image Library for experiments.The FCM algorithm optimized by other metaheuristic algorithms is compared with the original FCM algorithm to compare the accuracy and time-consuming of segmentation,and the time-consuming of each part of the code is recorded.The experimental results show that the most time-consuming part of the FCM algorithm iteration is to compute the image membership matrix.In the face of complex problems,the dragonfly algorithm can reduce the number of calculation times of membership matrix,find the global optimal solution faster,achieving optimization.By sacrificing a certain amount of precision,the algorithm greatly reduces time consumption and is suitable for the work that requires fast processing of complex images.
作者
黄卫
HUANG Wei(College of Artificial Intelligence and Computer,Jiang’nan University,Wuxi 214026,China)
出处
《应用科技》
CAS
2022年第5期32-38,共7页
Applied Science and Technology
关键词
图像分割
模糊C均值聚类
蜻蜓算法
元启发式算法
算法优化
隶属度划分
数据聚类
最大类间方差
image segmentation
fuzzy C-means clustering
dragonfly algorithm
meta heuristic algorithm
algorithm optimization
membership division
data clustering
maximum interclass variance
