一种新的交通流视频检测方法

提出一种基于彩色虚拟检测线的交通流信息视频检测方法．该方法通过分析光照、车道和阴影等模型之间的相互关系来判断车辆的存在，提高车流量检测的性能．通过引入色彩饱和度信息、自适应背景更新和运动检测，有效地区分阴影和车辆；通过二叉决策树来分析车辆的压线过程，提高了车流量检测的可靠性；通过一维视频跟踪的方式，解决了车辆速度检测的难题．试验结果表明，所采用的局部区域检测方法大大提高了交通流信息检测的实时性，且车流量检测的准确率可提高到98％．

New Video Detection Method for Traffic Flow

A new method using virtual trapline is proposed for traffic flow detection. For improving the detection accuracy, the system sets up several models, such as road model, illumination model, shadow model and relationship model among them to assist detection process. Through introducing the hue saturation degree information, adaptive background updating and motion detection, vehicles can be distinguished from shadow. Binary decision tree is used to analyze pressing line of vehicles to improve the reliability of the system. The problem of the vehicle velocity detection is resolved by one dimension video tracing. The results indicate that the new method is indeed better than existing methods in accuracy rate (can be improved to 98%) and real-time performance.