一种结合Grabcut的Vibe目标检测算法  

A Vibe moving objects detection combined with Grabcut

作　　者：邱祯艳[1] 王修晖[1] 

机构地区：[1]中国计量学院信息工程学院,浙江杭州310018

出　　处：《中国计量学院学报》2012年第3期250-256,288,共8页Journal of China Jiliang University

基　　金：浙江省自然科学基金资助项目(No.Y107003)

摘　　要：针对传统背景建模方法不能非常有效地提取出完整运动目标这一问题,提出了一种新颖的结合Grabcut和Vibe算法的运动目标检测方法.首先利用改进的Vibe目标检测算法初步寻找出当前帧中的所有可能目标前景区域;然后对这些前景区域进行进一步的选择从而滤去无效前景,合并有效区域;最后对每个前景区域再利用Grabcut算法来分割出更加完整的目标前景.实验结果表明,与传统的背景建模方法相比,本算法在有效时间内可以更加完整地分割出运动目标前景,对与背景颜色类似的目标前景也有很好的检测效果.To solve the problem that traditional background models cannot effectively extract complete moving objects,a novel strategy based on Vibe moving objects detection and Grabcut was proposed.Firstly,improved Vibe algorithm preliminary searches were used for all possible foreground target regions in the current frame.Secondly,those regions via abandoning invalid areas and merging valid areas.Lastly,by utilizing Grabcut algorithm the objects could be perfectly segmented from every region.Experiments showed that,compared with traditional methods,the algorithm can obtain more exact moving objects in effective time and has a better result even with background color that is greatly close to background.

关 键 词：运动检测 背景建模 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]