【作者】 龙海侠；

【导师】 须文波；

【作者基本信息】 江南大学 ， 计算机软件与理论， 2007， 硕士

【摘要】 聚类算法在数据分析,数据挖掘等许多地方有广泛的应用,该文探索了基于量子行为的微粒群优化算法(QPSO)的数据聚类及其在图像分割中的应用。首先,在分析K-Means聚类、PSO聚类、K-Means和PSO混合聚类(KPSO)的基础上,提出了基于量子行为的微粒群优化算法(QPSO)的数据聚类,并且研究了使用K-Means聚类的结果重新初始化粒子群,结合QPSO算法聚类,即KQPSO,介绍了如何利用上述的算法去找到用户指定的聚类个数的聚类中心,聚类过程是根据数据之间的Euclidean(欧几里得的)距离,K-Means算法、PSO算法和QPSO算法的不同在于聚类中心向量的“进化”上,使用了三个数据集比较了上面提到的五种聚类方法的性能,结果显示了基于QPSO算法的数据聚类的性能比较优越。其次,研究了基于QPSO的图像颜色分割方法,把图像分割问题看作一个最优化问题并且采用QPSO的进化策略聚类颜色特征空间中的区域,文中给出了三幅图像的分割效果,证明了QPSO算法在自动的和无监督的颜色分割上具有很好的效能。在QPSO算法中,收缩-扩张系数对于QPSO中的单个粒子的收敛来说是一个至关重要的参数。在文中使用了适应性机制,对数据聚类使用了适应性的基于量子行为的微粒群优化算法(AQPSO)。最后,该文使用一种新的距离度量方法进行聚类,实验证明了新的度量方法比Euclidean标准更具有健壮性,聚类的结果更精确。在此基础上使用QPSO算法进行数据聚类和图像分割,实验结果证明了QPSO算法优于PSO算法。QPSO算法不仅参数个数少,随机性强,并且能覆盖所有解空间,保证算法的全局收敛。更多还原

【Abstract】 Clustering algorithm has a wide of applications in many of fields, for example data analysis and data mining technology. Data clustering and the application in image color segmentation are explored using Quantum-behaved Particle Swarm Optimization (QPSO) in this paper.Firstly, QPSO algorithm is proposed to cluster data based on the K-Means clustering、PSO clustering and KPSO clustering and K-Means clustering is used to seed the initial swarm,combing with QPSO to cluster data,namely KQPSO. Introducing how to use these algorithms to find the centroids of cluster which a user specified number. All the process of clustering base on the Euclidean distance among data vectors. The difference between K-Means、PSO、QPSO is the evolution of the cluster-centroids. The performance of the five clustering method are compared on three data sets. The experiments results show QPSO clustering superiority.Secondly, image color segmentation is researched using QPSO algorithm. The problem of image color segmentation is regarded as an optimization problem and is adopted evolutionary strategy of QPSO for the clustering of regions in color feature. Three images results of segmentation are presented and demonstrate the efficiency of QPSO algorithms to automatic and unsupervised color segmentation.In QPSO, Contraction-Expansion Coefficient is a vital parameter to the convergence of the individual particle. Adaptive mechanism is used in this paper, therefore Adaptive Quantum-behaved Particle Swarm Optimization (AQPSO) is adopted to cluster date.Finally, new distance metric is used in clustering procedures. Experiment results show that this new metric is more robust and accuracy than common-used Euclidean norm. Using QPSO algorithm to data clustering and image segmentation based on the new metric .The experiment results show that QPSO is superior to PSO. Not only parameter of QPSO is few and randomicity of QPSO is strong, but also QPSO cover with all solution space and guarantee global convergence of algorithms.更多还原