自适应的over-relaxed快速动态均值漂移算法

来源期刊:中南大学学报(自然科学版)2008年第6期

论文作者:杨 斌 赵颖 樊晓平 周芳芳

文章页码:1296 - 1302

关键词:均值漂移;高斯核;边界优化;动态更新

Key words:mean shift; Gaussian kernel; bound optimization; dynamic update

摘    要:为了解决高斯核均值漂移算法收敛速度慢、计算效率不高的问题,提出自适应over-relaxed快速动态更新方法改进高斯核均值漂移算法。首先,在静态均值漂移算法中引入数据集的动态更新机制,每次迭代后将数据集更新到新的数据点,然后,将迭代过程中聚集在一起的数据点用1个收敛点表示,逐步减少参与计算的数据,保证准确性的同时降低计算量。由于非正态分布的数据集动态更新时,主方向上的数据点的收敛速度较慢,采用over-relaxed的策略来提高主方向数据点的迭代步长,并根据数据集直径的变化,自适应地计算步长参数。实验结果表明,改进后的高斯核均值漂移算法以超线性的速度收敛,收敛点的应用降低了收敛过程中的计算量。

Abstract: An adaptive over-relaxed fast dynamic mean shift was proposed to speed up the convergence of Gaussian mean shift. Firstly, the convergence speed of Gaussian mean shift was improved by dynamically updating the data set and the number of the data set decreased using a special point replacing the points overlapped during the iterations. Secondly, as the data set didn’t follow the normal distribution, the direction of principal component converged more slowly than other directions. So the convergence of the data in the direction of principle component was accelerated by the over-relaxed strategy and the parameter was adaptively calculated by the diameter of the dynamic data set. The experimental results prove that the convergence speed of the proposed method is non-linear and the use of convergence points can lower the complicacy of process of iteration.

基金信息:国家自然科学基金资助项目

相关论文

  • 暂无!

相关知识点

  • 暂无!

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号