像素域运动对象提取算法的研究  

Research on Moving Object Extraction from Video Pixel Domain

作　　者：周启亚 杨高波[2] 

机构地区：[1]湖南铁路科技职业技术学院信息技术系,湖南株洲412000 [2]湖南大学计算机与通信学院,湖南长沙412008

出　　处：《湖南工业大学学报》2008年第6期50-54,共5页Journal of Hunan University of Technology

基　　金：国家自然科学基金资助项目(60572127);湖南省自然科学基金资助项目(05JJ30113)

摘　　要：针对像素域运动对象的提取,较系统地分析比较了现有主要算法各自的原理、提取效果及时间特性,为研究者在各种不同的场合选择不同的提取算法提供了依据。为使算法具有可比性,实验在同一Hall_Monitor序列上进行。然后,提出一种改进的、同时进行帧间差分和减背景的运动对象提取算法,其鲁棒性在于能对光照变化、运动物体的暂停,显露部分的背景区域等复杂情况作出正确判断处理。In view of extraction process, the principles, the extracted effects and extracted time attributions of the main algorithms about the moving object extraction （ MOE ） from video pixel domain are analyzed and compared. The contribulions lie in the selection rules of the appropriation algorithm for own uses according to different conditions. Simulation results are demonstrated, which are conducted on Hall-Monitor sequence. Then, a improved MOE algorithm is presented, which is based on frame difference and background subtraction simultaneously. Its robustness can deal with some complicated situations such as light change, the paused moving object in background and the uncovered background.

关 键 词：运动对象 提取 视频处理 像素域 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]