一种新的背景模型建立及更新方法  

A Novel Method for Background Modeling and Update

作　　者：王华伟[1] 吴琦颖[1] 白建华[2] 李翠华[1] 

机构地区：[1]厦门大学计算机科学系 [2]厦门大学软件学院,福建厦门361005

出　　处：《厦门大学学报（自然科学版）》2005年第6期774-777,共4页Journal of Xiamen University：Natural Science

基　　金：国家自然科学基金(60175008)资助

摘　　要：提出一种新的建立背景模型和更新背景的方法.首先采用一种新颖的类Hausdorff距离建模方法建立初始背景,然后对每一输入帧使用背景差分法获得前景,并利用高效分类器创造性地将前景分类为运动目标、静止目标、虚假目标、噪声,最后在背景更新阶段使用基于前景目标的背景更新方法.实验结果表明,本方法能建立可靠初始背景,并能有效地解决背景更新“死锁”问题,增强背景模型的鲁棒性.This paper proposes a novel background modeling and update algorithm. Firstly a novel Hausdorff-like background modeling is built and for each image sequence foreground is obtained by background subtraction, then a foreground object classifier is used to get moving object,static object, false object, and noise, finally the background model is updated by a foreground object classification based background update algorithm. Extensive experiments results on indoor and outdoor image sequences demonstrate that the proposed system can effectively build a reliable background model and resolve the deadlock problem which results from false object.

关 键 词：背景模型 前景分类 虚假目标 背景更新 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]