摘要： 提出了一种在交通场景中鲁棒地检测运动车辆的方法，该方法采用改进的加权平均法进行实时背景学习，并根据统计量动态地计算出运动分割的阈值，以得到交通场景中的运动物体。与以往的方法相比，该方法在检测的成功率以及运算速度上均具有明显的优势

关键词: 视频处理；交通流量监测；背景学习；运动分割

Abstract: A method is presented to robustly detect the moving vehicles in traffic scenes. In the proposed method, the static background is subtracted in real-time with an improved weighted average algorithm. Then the threshold to do the segmentation is calculated dynamically,considering the variation of the illuminant. With such a dynamic local thresholds, moving objects are segmented by a differencing image scheme. Compared with the existed schemes, the proposed method performs better in recognition rates and the processing speed.

Key words: Video process; Traffic flow surveillance; Background learning; Motion segmentation