群體智能的概念來自於對群聚性生物的觀察研究，近年來已有許多不同演算法的研究，其中粒子群優化演算法(Particle Swarm optimization，PSO)演算法便是其中之一，由於具有設定參數少、有較好的全局搜索能力、搜尋結果較為穩定、應用範圍廣泛等特性，近年來被許多學者專家所研究，衍生出許多其它相關演算法，標準的PSO演算法雖有較好的全局搜索能力，但是愈接近最佳解，收斂速度愈慢。而另一聚類分析演算法K-means，實作簡單，收斂速度快，有較好的局部搜索能力，但是分類結果容易受初始設定值的影響，陷入局部最佳解的情形。這兩種演算法應用在圖像分割上，各有利弊，但是若結合在一起便能達到互補的效果，本研究便結合這兩種演算法的優點對圖像進行分割，並對實驗結果加以分析討論，結果證明，結合兩種演算法能比單獨使用PSO演算法有更快的收斂速度，並修正初始值對K-means演算法的影響。Abstract The concept of swarm intelligence gathered from the biological observation. In recent years there have been many studies, particle swarm optimization(PSO) algorithm is one of them, as a set few parameters, better global search capability, search results more stable and widely used. In recent years, many scholars and experts have researched PSO and derived many other related algorithms. Although the standard PSO algorithm has better global search capability, but the closer the optimal solution, convergence rate slower. The other clustering algorithm K-means, easily implement, rapid convergence, better local search capabilities. But the result of K-means heavily depends on initial cluster centers. It may find the local minima if initial cluster centers are not good. This research combines the advantages of PSO and K-means for image segmentation. Results show that a combination of two algorithms can be convergence faster than PSO and fix initial problem of K-means algorithm.
