基于监控视频的前景目标提取  

Target extraction based on monitoring video

作　　者：雷瀚清 李豪 滕国伟[1] Lei Hanqing;Li Hao;Teng Guowei(College of Communication, Shanghai University, Shanghai 200444, Chin)

机构地区：[1]上海大学通信学院

出　　处：《电子测量技术》2018年第7期64-68,共5页Electronic Measurement Technology

摘　　要：前景目标提取是智能监控视频的基础部分。针对监控视频稳定和晃动两种情况下前景目标提取不理想的现象,提出了基于分段加权加滑动平均背景建模和菱形搜索算法的监控视频前景目标提取方法,基于分段加权加滑动平均背景建模的前景提取方法,通过分割历史像素值,使用加权平均的方法计算分割数据段的长度和均值,进而建立背景模型,利用背景减除的方式提取前景目标。而菱形搜索算法能有效地计算出晃动视频的全局位移矢量MV,利用MV可以得到晃动视频中的前景目标图像。实验表明,基于分段加权加滑动平均背景建模的前景提取方法能很好地克服不晃动视频中前景目标提取不完整的缺点,提取精度有所上升,菱形搜索算法对于晃动视频中的运动目标提取效果噪点少,实时跟踪效果得到改善。The foreground target extraction is the basic part of intelligent monitoring video.For monitoring and video stable rock in both cases the phenomenon of foreground object extraction is not ideal,is presented based on piecewise weighted moving average background modeling and prospect of surveillance video object extraction method of diamond search algorithm,based on the weighted moving average background modeling method to extract the prospect of history by splitting pixel values,using the weighted average method to calculate length and the mean of the data,and the background model,foreground target is extracted using background deduction.The rhombic search algorithm can effectively calculate the global displacement vector MV of the shaking video,and the image of foreground target in video can be obtained by using MV.Experiments show that,based on the weighted moving average background modeling the prospect of extraction method can overcome not shake well prospect in the video object extraction incomplete fault,extraction accuracy has increased,the diamond search algorithm for moving object in video shaking extraction effect less noise,real-time tracking effect is improved.

关 键 词：前景提取 背景减除 背景建模 菱形搜索 

分 类 号：TN919[电子电信—通信与信息系统]