【作者】 胡萍萍；

【导师】 裴振奎；

【作者基本信息】 中国石油大学 ， 计算机软件与理论， 2008， 硕士

【摘要】 微粒群算法(PSO, Particle Swarm Optimization)是一种新近出现的启发式全局优化算法,由于算法的易实现性和高效性,因此受到了人们的广泛关注。它已成为与遗传算法、禁忌搜索算法以及模拟退火算法并行发展的一种全局优化算法。这种算法的基本思想是源于鸟类的群体行为。其特点是在较为简单的规则指导下,可以保证微粒迅速降落在最优解处。因此,这种方法具有一定的智能性和社会性。基于PSO的基本思想,本文对PSO算法进行了详细的论述,讨论了算法的优缺点,并对PSO的基本原理进行了详细分析。为了进一步改进算法的性能,本文提出了两种新算法。一种是基于PSO的核模糊聚类算法(Kernel Fuzzy C-means clustering)。通常,传统聚类算法只在数据特征差异较大时才有效,当数据特征差异较小时,传统聚类算法很难取得较好的聚类效果。新算法先用高斯核函数,把输入空间的样本映射到高维特征空间后,再利用微粒群算法的全局搜索、快速收敛的特点,在特征空间中进行聚类,克服了KFCM对初始值和噪声数据敏感、易陷入局部最优的缺点。通过对医学图像的分割,结果表明,新算法在性能上比KFCM聚类算法有较大的改进,具有更好的聚类效果,且算法能够很快地收敛。另一种是基于PSO与DCT域的鲁棒图像水印算法。数字水印技术是保护多媒体数据版权和图像可靠性认证的一种新技术,本文比较系统地研究了数字水印在图像中的应用问题,提出了一种新的基于PSO与DCT域的鲁棒图像水印方案。实验仿真表明,针对不同的水印图像和不同的载体图像存在不同的优化频带,在优化频带中嵌入水印明显地平衡了鲁棒性和隐蔽性的冲突。攻击实验表明,该算法对压缩、噪声等一般信息处理是鲁棒的。本文还讨论了置乱技术在数字水印中的应用,置乱技术能分散错误比特的分布,提高了数字水印的鲁棒性。最后,简述了PSO算法在未来在图像处理领域的发展前景及相关应用,据此指出了未来的主要研究工作和方向。更多还原

【Abstract】 Particle swarm optimization algorithm is a heuristic global optimization algorithm which appeared recently. It has been widely concerned by people because of its feasibility and effectiveness. It has been proven to be a powerful competitor to other heuristic algorithms, such as genetic algorithm, taboo search and simulated annealing algorithm for global optimization problems. The base idea of this theory is from the colony behaviors of birds. The merit of PSO is that it can assure the particles land the best place with some simple rules. So this method has the attribute of intelligence and society partly.Based on the PSO’s idea, the PSO has been introduced and discussed in detail in this paper. The merits and disadvantages of the PSO are analyzed and then its basal principle is analyzed and researched in this paper. In order to prove algorithm’s performance, two novel algorithms are improved. One is Fuzzy clustering algorithm which uses the merits of the global optimizing and higher convergent speed of Particle Swarm Optimization (PSO) algorithm and combines with Kernel Fuzzy C-Means (KFCM) is proposed. In general, traditional clustering algorithms are suitable to implement clustering only if the feature differences of data are large. If the feature differences are small and even cross in the original space, it is difficult for traditional algorithms to cluster correctly. By using Gauss kernel functions, we can map the data in the original space to a high dimensional feature space in which we can perform clustering efficiently. The algorithm eliminates KFCM trapped local optimum, being sensitive to initial data and the noise data. The performance of this modified PSO-KFCM is compared with KFCM. The results of simulation experiments on medical image show the feasibility and effectiveness of the new clustering algorithm. Another is watermarking scheme on PSO in the DCT domain. Digital watermarking has recently been proposed as a new means to provide copyright protection of multimedia data and image authentication. Along with penetrated analyze of the theory of PSO, an innovative watermarking scheme on PSO in the DCT domain is proposed. Computer simulation indicates this scheme can find an optimized DCT domain frequency bands when we embed an watermarking image into an digital image. Simulation results prove the balance of conflicting between the invisibility and robustness. Attacking experiment results reveal the fact that the scheme is robust to JPEG compression, noise, et al.Finally, the developing foreground and correlative application engineering technology are introduced. The research trend and PSO in the future are pointed out in some areas.更多还原