改进的均值漂移算法在运动目标跟踪中的研究  

Research of Improved Mean Shift Algorithm in Moving Object Tracking

作　　者：王娟[1,2] 刘教民[2] 孟军英[2] 韩明[2] 

机构地区：[1]中国人民武装警察部队学院,廊坊065000 [2]燕山大学信息科学与工程学院,秦皇岛066004

出　　处：《系统仿真学报》2012年第9期1896-1899,共4页Journal of System Simulation

基　　金：河北省自然基金(F2012208004)

摘　　要：以智能交通系统中运动目标跟踪为背景,研究采用均值漂移算法,实现对快速移动的单目标事物准确实时跟踪。首先,分析了均值漂移算法的基本原理。其次,分析了均值漂移算法在目标跟踪中存在的问题及解决方法:针对目标遮挡情况下目标跟踪的准确性问题,采用将均值漂移算法与卡尔曼滤波算法相结合的自适应均值漂移目标跟踪算法;针对大小不变的带宽窗口可能导致跟踪丢失的问题,采用了自动调节带宽窗口的方法。最后进行了仿真实验。Fast moving object was accurately tracked based on the intelligent transportation system. Firstly, the principle of mean shift algorithm was analyzed. Then, the mean shift algorithm in target tracking problems and solutions were analyzed： target under occlusion tracking accuracy, the mean shift algorithm and Kalman algorithm combining adaptive mean shift tracking algorithm; size constant bandwidth window leading to loss of tracking problem; automatic adjusting the bandwidth of the window method. Finally, through the simulation experiments, the tracking effects were compared.

关 键 词：均值漂移 目标跟踪 智能交通 卡尔曼滤波 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]