融合帧差和ViBe的运动目标检测算法  

Moving object detection algorithm using ViBe combined with frame-difference

作　　者：魏洪涛[1] 李瑾[1] 吴细秀[2] 

机构地区：[1]武汉理工大学信息工程学院,武汉430070 [2]武汉理工大学自动化学院,武汉430070

出　　处：《计算机应用研究》2017年第5期1565-1567,1571,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(51107093)

摘　　要：ViBe算法是一种快速高效的背景建模算法,但该算法在运动目标检测过程中会产生鬼影。针对ViBe算法中鬼影消除缓慢的问题,结合多个场景的交通视频提出一种通过连续两帧前景背景像素时域变化来判断鬼影像素点并消除的方法,该方法加快了鬼影的消除速度。同时,对于视频拍摄场景中的背景噪声,采用了对前景图进行开闭操作去除小像素点以及对目标区域的空洞进行填充处理。实验表明,改进的ViBe算法能够加快鬼影的消除,并且与帧差法以及混合高斯建模算法相比,前景检测效果更精确。ViBe algorithm is an effective and quick background modeling algorithm. During the moving target detection,the ViBe will produce the ghost and eliminate slowly. In order to solve the problems caused by the execution of ViBe algorithm,this paper proposed an improved algorithm for the traffic video. It judged the ghost pixel by the change of two consecutive frames in time domain,which could speed up the ghost to eliminate. In addition,to solve the noise in the complex background,it applied a morphology post-processing with opening operation and closing operation to remove the noise,then filling the holes in the target zone to highlight the prospect target. Experiments show that the improved algorithm can absorb the ghost quickly and detect the foreground better compared with frame-difference and GMM.

关 键 词：运动目标检测 ViBe算法 帧差法 鬼影消除 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]