基于粒子滤波和Mean_shift的跟踪算法.pdf

第36 卷 第5 期 计 算 机 工 程 2010 年3 月 Vol.36 No.5 Computer Engineering March 2010 ·博士论文· 文章编号：1000—3428(2010)05—0021—02 文献标识码：A 中图分类号：TP391.41 基于粒子滤波和Mean-shift 的跟踪算法 1 1 2 1 蒋 旻 ，许 勤 ，尚 涛 ，高伟义 (1. 武汉科技大学计算机科学与技术学院，武汉 430081 ；2. 武汉科技大学信息科学与工程学院，武汉 430081) 摘 要：粒子滤波作为一种基于贝叶斯估计的算法，在处理非线性运动目标跟踪问题上具有特殊的优势。基于此，提出一种基于粒子滤波 和Mean-shift 的混合跟踪算法(KMSEPF) 。KMSEPF 算法对一般的Mean-shift 和粒子滤波混合算法进行改进。结果证明，KMSEPF 算法与 混合算法MSEPF 相比，在计算效率提高的同时，跟踪准确性和处理遮挡的能力没有下降。 关键词：粒子滤波；Mean-shift 算法；目标跟踪 Tracking Algorithm Based on Particle Filtering and Mean-shift 1 1 2 1 JIANG Min , XU Qin , SHANG Tao , GAO Wei-yi (1. College of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan 430081; 2. College of Information Science and Engineering, Wuhan University of Science and Technology, Wuhan 430081) 【Abstract 】As an algorithm based on Bayesian estimation, particle filtering is predominant on tracking nonlinear moving target. This paper proposes an algorithm, which is based on Mean-shift and particle filtering, named K-means and Mean-shift Embedded Particle Filter(KMSEPF). The KMSEPF algorithm improves the general mixture algorithms which are based on particle filtering and Mean-shift. Results show that the algorithm reduces the computation complexity, while maintains the high precision and the ability to control the occlusion, compared with the MSEPF algorithm. 【Key words 】particle filtering; Mean-shift algorithm; object tracking 1 概述