Abstract: For many uncertain factors of intersection video surveillance such as more vehicles，uncertain speed and the changing environment and so on.So it is difficult to detect moving objects efficiently from video sequences.W4 algorithm is used in intersection video surveillance system.W4 algorithm first uses a pixelwise median filter to distinguish the stationary pixels from the video and uses these stationary pixels to form the background parametric model.At the same time according to the conditions to choose the updating methods pixel-based or object-based.Experimental results show that the W⁴ algorithm possesses higher performance on intersection video surveillance under complex environments.

Key words: vehicle detection, W⁴ background model, intersection

摘要： 在智能交通视频监控系统中，由于交通路口车辆多、车速不定且外界环境光线变化等诸多因素造成不能很好地检测到运动目标。将W⁴算法应用到道路交叉口的视频监测中，该算法首先利用中值法进行初始化选出静止像素集合，并利用静止像素构建背景参数模型，同时建立基于像素级和基于目标级的背景更新模式。通过仿真验证该算法能够在路口比较复杂的情况下很好地检测到车辆，与传统的方法相比检测的准确度有了较大的改进。

关键词: 目标检测, W⁴背景建模, 道路交叉口