一种持续光流跟踪的运动目标检测方法  

An Approach for Moving Object Detection Using Continuing Tracking Optical Flow

作　　者：石祥滨[1] 王萌[1] 张德园[1] 黄小水 

机构地区：[1]沈阳航空航天大学计算机学院,沈阳110136

出　　处：《小型微型计算机系统》2014年第3期642-647,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61170185)资助;辽宁省科技攻关计划项目(2011217002)资助;辽宁省博士启动基金项目(20121034)资助

摘　　要：针对摄像机与被检测目标同时运动情况下的目标检测问题,提出一种适合于序列图像的持续光流跟踪的运动目标检测方法.首先,在起始两帧间使用金字塔稀疏光流法跟踪特征点,将跟踪到的特征点根据其运动进行分组,然后在后续帧图像上持续跟踪已分组的特征点,对于每组中的特征点,只保留持续跟踪到的特征点最大集合,去除过小分组,直至筛选出只属于背景的特征点分组用于背景运动补偿,最后使用帧间差分法检测运动目标.对于后续帧,仅在已筛选出的背景特征点分组上持续光流跟踪.实验证明,本文提出的方法能够提高检测运动目标的精度和速度.To aim at detecting object when the camera and object motions are mixed together, this paper proposes an approach for moving object detection using continuing tracking optical flow in image sequences. First, it tracks feature points between first and sec- ond image frames using the pyramidal Lucas-Kanade method and then groups these tracked points based on their motions. Next, it continues tracking the grouped feature points on subsequent image frames. For these points in each group, the maximum aggregate will be reserved and undersize groups must be removed in the continuing tracking process, until the only group is to be chosen. All the feature points belong to background in this group and it can be applied to compensate for background motion. Finally, it detects the moving object employing frame difference method. For subsequent image frames, the feature points in the only group are just con- tinued tracking. The experiment results demonstrate that the method put forward in this paper improves the accuracy and the speed of detecting moving object.

关 键 词：运动目标检测 摄像机运动 持续光流跟踪 背景运动补偿 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]