应用模拟退火粒子群算法优化二维熵图像分割  

2-D entropy image segmentation based on simulated annealing particle swarm optimization

作　　者：吴禄慎[1] 程伟 王晓辉[1] WU Lu-shen;CHENG Wei;WANG Xiao-hui(School of Mechanical and Electrical Engineering,Nanchang University,Nanchang 330031,China)

机构地区：[1]南昌大学机电工程学院

出　　处：《计算机工程与设计》2019年第9期2544-2551,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(51365037)

摘　　要：针对二维熵图像分割方法存在计算时间长、实时性差的问题,提出一种基于模拟退火粒子群算法的二维熵图像分割方法。将模拟退火机制引入粒子群算法(PSO),根据初始种群的最优适应度值设置初始温度,采用Metropolis准则优化生成个体最优位置和全局最优位置,对PSO算法的惯性权重参数进行优化,避免粒子在寻优过程中陷入局部最优,提高算法的收敛速度。对多幅具有不同直方图分布的图像进行阈值分割实验,实验结果表明,该方法与二维熵穷举分割法相比,分割结果相同,分割效率大大提高,分割阈值选取的准确性和运算时间都优于遗传算法(GA)和PSO算法。To overcome the problems of long computing time and poor real-time performance in 2-D entropy image segmentation,an 2-D entropy image segmentation based on simulated annealing particle swarm optimization was proposed.To prevent particles from falling into local optimal in the process of optimization and to improve the convergence speed,simulated annealing mechanism was introduced into particle swarm optimization(PSO).The initial temperature was set up according to the optimal fitness value of the initial population.The Metropolis rule was used to optimize personal best position and global best position and to optimize the inertia weight factor of PSO.Different histogram distributions of image segmentation experiments were performed.Experimental results show that compared with 2-D entropy exhaustive segmentation method,the proposed method can get the same results,and the running speed is greatly improved.Moreover,the threshold selection accuracy and running speed of the proposed method are both better than that of GA and PSO.

关 键 词：图像分割 二维熵 粒子群算法 模拟退火 惯性权重 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]