基于运动相机的运动目标检测方法研究

Research of moving target detection method based on moving camera

随着视频监控的广泛应用,为对运动目标进行更好的监控,相机被搭载在移动平台上,因此研究基于运动相机的运动目标检测方法具有重要的意义。从研究视频运动目标检测技术开始,包括帧间差分法、背景差分法、光流法3类算法,具体对差分法、单高斯建模、高斯背景建模、H-S光流法、L-K光流法进行了原理研究。采用OpenCV技术对帧差法、高斯建模、H-S光流法进行仿真实验,总结算法的优点和缺点。最后针对H-S光流矢量与目标运动特性的相关性,提出了一种基于H-S光流的运动目标检测方案。

With the wide application of video surveillance,camera is mounted on the mobile platform in order to better monitor the moving target.Therefore,it is of great significance to study the moving target detection method based on moving camera.This paper first studies the video moving target detection technology,including inter-frame difference method,background subtraction method and optical flow method.In this paper,the principle of inter-frame difference method,basic Gaussian modeling,mixture Gaussian modeling,H-S optical flow method and L-K optical flow method are studied.The OpenCV technique is used to simulate the inter-frame difference method,Gaussian modeling and H-S optical flow method,Summarize strengths and weaknesses.Finally,aiming at the correlation between H-S optical flow vector and target motion characteristics,a moving object detection scheme based on H-S optical flow is proposed.