基于多区域的人体运动跟踪研究与应用  

Research on Human Motion Tracking Based on Multi-Region and its Application

作　　者：高书秀[1] 黄剑华[1] 唐降龙[1] 刘家锋[1] 刘国军[1] 

机构地区：[1]哈尔滨工业大学计算机学院,150001

出　　处：《微计算机信息》2008年第19期267-269,共3页Control & Automation

基　　金：国家自然科学基金(60672090/F010204);移动拍摄下MCMC粒子滤波的人体运动跟踪算法研究

摘　　要：针对单区域跟踪快速运动人体易产生漂移的缺点,本文提出一种基于Kalman Filter预测的多区域跟踪的新算法。该算法利用Kalman Filter预测人体各区域,然后利用颜色直方图匹配算法精确定位人体各区域,最后加权计算人体最终位置。Most signal-region tracking algorithms track fast moving players that occurs excursion, then a novel approach integrating multi-region into Kalman Filter is proposed to track fast moving players in sport. First, the predicted position of each region is calculated using Kalman Filter. Second, the weight of each region on the fixed-grid sample is obtained by kernel-based color histogram matching algorithm. Last, optimal estimated position of each region in current frame is the weighted expectation calculated by the weight of each region.

关 键 词：KALMAN FILTER 多区域跟踪 颜色直方图 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]