基于ViBe的改进对运动目标边缘检测  

Edge detection of moving target based on improved ViBe

作　　者：张晗 钱育蓉[1] 王跃飞[1] 陈人和 田宸玮 ZHANG Han;QIAN Yu-rong;WANG Yue-fei;CHEN Ren-he;TIAN Chen-wei(School of Software,Xinjiang University,Urumqi 830008,China)

机构地区：[1]新疆大学软件学院

出　　处：《计算机工程与设计》2019年第9期2471-2477,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61562086、61462079);新疆维吾尔自治区创新团队基金项目(XJEDU2017T002)

摘　　要：为提高视频中运动目标边缘检测的精度并消除鬼影,从运动区域标记上,提出构建部分邻域背景模型,避免检测到过多无关信息的思想;从运动目标边缘检测上,提出从部分邻域和全域中抽样的方法,用二者概率绝对差值作为边缘检测的依据。经测试,改进后算法实现了鬼影的消除,避免检测到过多大面积无关背景边缘,在小于1024*1024分辨率上可获得较好的处理速度。该方案易于实现,能够加快鬼影消除,与帧差法融合使用可使帧差法检测效果更精确。To improve the accuracy of moving target edge detection in video and eliminate ghosting,from the perspective of the motion region labeling,the idea of constructing a partial neighborhood background model to avoid detecting too much extraneous information was proposed.From the perspective of the detection of moving target edges,a method of sampling from partial neighborhood and global domain was proposed.The difference was used as the basis for edge detection.After testing,the improved algorithm achieves the elimination of ghosts,and avoids detecting too many large-area irrelevant background edges.A higher processing speed can be obtained at less than 1024*1024 resolution.The scheme is easy to be implement,and it can accelerate ghost elimination,and can be combined with the frame difference method to make the frame difference detection effects more accurate.

关 键 词：运动目标边缘检测 ViBe算法 鬼影消除 部分邻域背景模型 帧差法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]