基于改进Camshift算法的视频对象跟踪方法  

An Object Tracking Method Based on Improved Camshift Algorithm

作　　者：刘雪[1] 常发亮[1] 王华杰[1] 

机构地区：[1]山东大学控制科学与工程学院,山东省250061

出　　处：《微计算机信息》2007年第21期297-298,305,共3页Control & Automation

基　　金：山东省自然科学基金重点资助项目(Z2005G03)

摘　　要：针对Camshift算法限于简单背景下跟踪特定颜色目标的局限性,提出一种改进的Camshift算法。首先在HSV色彩空间建立目标的背景加权三维直方图,将视频图像转化成颜色概率图,然后结合颜色概率分布图与差分图像,根据目标运动速率大小自适应的赋予两者不同的权重,实现了复杂背景下的目标跟踪。实验结果表明,该方法是十分有效的。Considering the localization of Camshift algorithm which only applies to track targets with certain color in simple back- ground, an improved Camshift algorithm is proposed. First three dimensional background weighted histogram is built in HSV color space, and video image is converted into color probability distribution image. Then combine color probability distribution image with differential image by using different weights adaptively according to the target velocity. So tracking targets in complex situation is realized. Experimental results show good performances of the proposed algorithm.

关 键 词：目标跟踪 CAMSHIFT 背景加权直方图 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]