中文摘要：

为了提高交通监控的准确性,针对传统ViBe算法存在鬼影问题,提出了一种V-ViBe算法的改进方案。该算法利用连续的图像序列代替单帧图像,构造出＂虚拟＂背景图像,改进了传统ViBe算法初始背景模型建立方法。利用形态学相关知识消除干扰目标,使检测目标更加完整。实验表明：V-ViBe算法从第一帧即可有效消除鬼影对目标车辆检测的影响,在不改变ViBe算法优势的前提下,显著提高了检测的准确性。

英文摘要：

In order to improve the accuracy of traffic monitoring,a V-ViBe（ Virtual background based ViBe） algorithm is proposed to deal with the ghost problem existing in classic ViBe algorithm.The algorithm constructs＂virtual＂background images by using continuous image sequence instead of a single frame image,which makes improvement for initial background modeling in traditional ViBe algorithm. Meanwhile,the morphological knowledge is utilized to remove the effect of interference targets so as to detect a more complete target. The experimental results show that the V-ViBe algorithm can effectively eliminate the influence of ghost on to target vehicle detection from the first frame,and the accuracy of detection is improved significantly without changing the advantages of ViBe algorithm.