基于视频流连续性的快速运动目标检测方法  

Fast moving object detecting method based on video streaming continuity

作　　者：张娟[1] 毛晓波[1] 陈铁军[1] 吴敏[1] 

机构地区：[1]郑州大学电气工程学院,郑州450001

出　　处：《计算机应用研究》2010年第8期3198-3200,共3页Application Research of Computers

基　　金：河南省自然科学基金资助项目(0611052900);河南省教育厅自然科学研究计(2009A410001);河南省留学人员科技活动择优资助基金资助项目(豫人留学函[2007]17号)

摘　　要：提出了一种基于视频流连续性的快速运动目标检测算法。对于背景模型的建立,在原有像素归类背景重构算法的基础上,基于视频流的连续性,选择性使用训练阶段背景至少在50%的时间内可以被观测到的前提假设,提出了一种改进的快速背景重构算法,在准确建立背景模型的同时,有效地减少了运算时间。对于运动阴影的干扰,采用基于RGB色度比例的色度差分算法,能快速有效地去除运动阴影,最终实现运动目标的快速提取。实验结果证实了所提算法的有效性。This paper proposed a fast moving object detecting algorithm based on the quality of video streaming continuity. In the aspect of background reconstruction,introduced a fast background reconstruction method. It was based on pixel intensity classification method and the assumption of background pixels could be observed in at least 50% time in the training phases. In the aspect of shadow removing,presented RGB chromaticity difference algorithm based on RGB chromaticity ratio. It could remove the shadow of moving object quickly and effectively. The experimental results confirm the validity of the proposed method.

关 键 词：背景差分 背景重构 像素归类 阴影去除 视频流 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]