基于边缘的背景差法在车流量检测中的应用

Application of edge-based background difference in traffic volume extraction

针对视频车流量检测容易受车辆阴影和车辆变道影响的问题,笔者提出了一种基于边缘信息的背景差车流量检测方法。该方法利用边缘信息作为车辆的检测特征,实时自动提取和更新背景边缘并采用动态开窗的方式来进行车辆计数。实验结果表明,与传统背景差法相比,该方法受车辆阴影和车辆变道影响较小,检测准确率达97.3%,是一种实用有效的车流量检测方法。

Video-based traffic volume extraction systems were easy to be influenced by vehicle shadow and vehicle lane-change. To solve these problems, a traffic volume extraction method using edge-based background difference was presented, This method used edge information as detecting characteristic of the vehicles, extracted and updated the edges of background automatically and counted vehicles by creating windows dynamically instead of using fixed windows. The experiment result showed that the proposed method is influenced less by vehicle shadow and vehicle lane-change when compared with traditional background difference method, has a detecting accuracy of 97.3%, and is an effective method to detect traffic volume.