基于视频分析的呼吸检测算法

摘要

针对传统视频监控蒙被子时无法进行呼吸判断的问题,提出一种改进的欧拉影像放大算法(Eulerian video magnification,EVM)进行呼吸检测,并结合运动历史图(Motion history image,MHI)和网格标记法来去除干扰影响.该算法核心是利用改进的EVM获得有运动前景目标,利用MHI来检测图像中的大运动,并用网格标记法除去手、头等外部运动干扰.结果表明:该方法实现简单,实时性好,可以有效除去呼吸以外运动对正常呼吸的干扰.%In the case of that in the traditional video surveillance,it is hard to judge wether the breathing under the quilt is normal or not,an improved Eulerian Video Magnification (EVM) is proposed to detect breathing in this paper.The Motion History Image (MHI) and marked grid method are combined with to remove interferences.The core of algorithm is to get foreground moving object with improved EVM,capture substantial motions with Motion History Image (MHI) and remove interferences coming from the movements of hands or head with marked grid method.Results show that the method is simple,has good real-time performance.It can effectively remove the interferences of other movements expect normal breathing.