基于RPCA与三帧差分融合的运动目标检测

为了实现在监控视频中对人体运动目标的准确提取,针对传统的三帧差分法在运动目标提取过程中容易出现“空洞”现象,提出了一种鲁棒主成分分析(Robust Principal Component Analysis,RPCA)与三帧差分相融合的运动目标检测算法。本算法通过将RPCA提取的视频当前帧的背景作为三帧差分法的中间帧与视频当前帧的前一帧和视频当前帧分别进行邻间差分,使得三帧差分法在运动目标检测过程中避免了背景像素点所带来的影响,消除了“空洞”现象。仿真结果表明该算法在完整性和准确性方面要优于其他三种传统运动目标检测算法,可以在复杂背景环境中实现准确的运动目标提取。

Moving object detection based on three frame difference fused with RPCA

To realize the accurate extraction of the moving object in the surveillance video,since the traditional three frame difference method can not avoid "hollow" in the process of moving object detection,this paper proposed a three frame difference method fused with Robust Principal Component Analysis(RPCA) .This algorithm works by using the background of the current frame of the video to difference with the previous frame of the video and the current frame of the video.It makes the three frame difference method avoid the influence of the background pixels in the process of moving object detection and eliminate the "hollow" phenomenon.The simulation results show that the proposed method performs better than the other algorithms and moving object can be detected correctly in complicated background circumstances.