基于复式滑动窗口的车辆排队长度检测方法

Measurement Method of Vehicle Queue Length Based on Duplex Sliding Window

随着生活水平的提高,居民拥车率逐渐上升,复杂的道路交通情况已难以用人力监控,智能交通系统可以很好地解决这一难题。车辆排队长度作为智能交通系统监控道路交通的基础参数之一,对交通流量的预测、交通信号灯的控制等起着关键性作用。文中基于视频图像处理,提出了一种固定窗口结合滑动窗口的车辆排队长度检测方法。使用位于队头的固定窗口准确描述车辆队列形成及消散的过程,通过复式滑动窗口实时跟踪排队车辆队尾位置确定排队长度。根据窗口区域内车辆是否存在与是否运动,构建队尾滑动窗口。实验结果表明,文中算法能实时准确地跟踪车辆队列队尾位置,可适应不同天气环境和光照变化,检测准确率达90%以上,为监控车辆堵塞和控制信号灯等提供了良好基础参数。

With the growing in the living standards of risidents, the number of automobiles are increasing rapidly, which makes it more and more complex to monitor the heavier traffic. However, this kind of issue can be solved effectively by intelligent transportation system. Being one of the significant parameters while monitoring the traffic in the system, vehicle queue length plays an important role in forecasting the traffic flow as well as controlling the traffic signal. Based on video image processing technology, an efficient vehicle queue method u- sing cooperation of fixed window and sliding window was proposed. The method can detect vehicle queue formation and dissipation with fixed window and track the position changes with duplex sliding window in real time. By vehicle presence detection and movement analysis on the monitored region, queue tall flexible window was adopted to track the queue tail position. The experimental results show that the accuracy rate of the presented algorism is over 90% by real-time tracking queue tall position under various illumination and weather conditions. Therefore, the proposed algorithm can provide the well fundamental parameters for traffic congestion monitoring and traffic signal controlling.