摘要
针对矿井危险区域视频监控视场背景复杂,难以实现视频目标精确提取的问题,提出了一种基于背景配准的视频目标检测算法。该算法实现步骤:提取SIFT特征点,计算特征点区域H-S光流矢量;通过区域运动特性分析提取出背景运动区域,对背景运动区域特征点做帧间匹配;计算仿射参数,配准差分后提取出精确的目标区域。实验结果表明,该算法能够去除前景目标特征点对背景配准的影响,可获得较为精确的目标区域。
In view of problem that background of field video of video monitoring in dangerous area of mine is complex,so as to be difficult to realize accurate extraction of video object,the paper proposed a video detection algorithm for object based on background registration.Implementing steps of the algorithm are as follows： extracting SIFT feature points and calculating H-S optical flow vector of the feature points area;extracting motion area of background by analyzing area motion performance and making frames matching for feature points of motion area of background;calculating affine parameters and extracting accurate object area after registration and difference.The experiment result showed that the algorithm can remove influence of feature points of foreground object on background registration and can gain accurate object area.
出处
《工矿自动化》
2011年第3期48-50,共3页
Journal Of Mine Automation
关键词
矿井危险区域
智能视频分析系统
视频目标检测
SIFT特征
H-S光流矢量
背景配准
仿射参数
dangerous area of mine
intelligent video analyzing system
video detection of object
SIFT feature
H-S optical flow vector
background registration
affine parameter
