一种基于视觉的车辆跟踪及交通流量参数估计新方法

New Method of Vision Based Vehicle Tracking and Traffic Parameter Estimation

交通监控系统采集处理有关道路车流状况的实时信息,以指导车辆的运行,保障道路顺畅。基于视觉的交监控系统具有直观明了、系统使用和维护费用相对较低等优点,因而可广泛应用于公路干线和交叉道口的交通监控。本文提出了一种改进的背景差法,改善了车辆检测效果,在此基础上进一步提出了用于灰度图像的综合车输检测方法,以消除阴影的影响,提高车辆检测的准确性,进而实现车辆的跟踪、计数、车型划分和行驶速度估计。实验所用视频图像信号中路边建筑造成的阴影从无到有,直至覆盖整个路面。实验结果显示该方法简单有效,能实时提取交通信息,其车辆检测的准确率达98%以上,优于背景差法和边缘检测法。

Vision based traffic monitoring systems have several apparent advantages such as easily intervened and lower costs. An improved background subtraction method together with a new method that combines the improved background subtraction with edge detection is proposed for vehicle detection and shadow rejection, based on which real-time vehicle trucking, counting, classification and speed estimation are implemented. Experiment demonstrated that the vehicle detection rate is larger than 98%, during which the passive shadows resulted from roadside buildings grew considerably.