改进粒子群结合K 均值聚类的图像分割算法

K 均值聚类的分类结果过分依赖于初始中心的选择且容易陷入局部最优。文中针对K 均值的缺陷，提出了一种基于随机权重粒子群和K 均值聚类的图像分割算法RWPSO KM。在算法开始，利用随机权重粒子群算法的全局搜索能力避免算法陷入局部最优。然后根据公式计算种群多样性执行K 均值算法，利用K 均值算法的局部搜索能力实现算法的快速收敛。实验结果表明, RWPSO KM与K 均值聚类和PSOK相比具有更好的分割效果和更高的分割效率。

A Modified PSO Combined K Means Clustering Algorithm for Image Segmentation

The heavy dependence of the K means clustering classification on selecting of the initial centers makes it easy to fall into local optimum. A RWPSO KM based on random weight particle swarm algorithm and K means algorithm is proposed. The global search capability of random weight particle swarm optimization algorithm is used first to avoid falling into local optimum, after which the population diversity is calculated according to the formula to execute K means algorithm, and the local search of K means algorithm is employed to achieve fast convergence. Experimental results show that RWPSO KM is superior to K means clustering and PSOK in segmentation effect and efficiency.