基于高斯混合模型的多车道车流量检测算法

Multilane Traffic Flow Detection Algorithm Based on Gaussian Mixture Models

研究视频的城市交通路口车流量检测准确率的问题,由于车速过慢,有效性差。针对目前的车流量检测算法仅限于单车道车流量检测及准确率低的问题,提出了基于高斯混合模型的多车道车流量检测算法,在交通路口的视频中设定所要检测多个车道的检测带并根据车道线划分车道,运用高斯混合模型对检测带进行背景建模,结合背景差法提取运动车辆,通过垂直投影方法解决车辆断层引起误检的问题,对车身宽度与阈值的比较判断车辆是否通过检测带,实现了多车道车流量检测。实验证明,多车道算法能有效克服断层引起误检的问题,检测车辆准确率高,实时性好,鲁棒性高,为智能交通灯控制提供准确参数。

Research detection rate of vehicle based on video of urban transportation crossroads.The existing traffic flow detection algorithms are limited to single lane and their accuracy is low.Multilane vehicle detection algorithm based on Gaussian Mixture Models was proposed.The algorithm,sets up detection region which coveres multilane and divides lane according to lane lines,uses Gaussian Mixture Models to build background model at detection region,combines background subtract method to separate moving vehicle,solves false detection problem caused by vehicle faultage by taking vertical projection,detects that vehicle through detection region according to the comparison result of moving object＇s width and threshold,and realizes multilane vehicle detection.The experiments indicate that the new method is useful to solve false detection problem and is better in real-time performance.What＇s more,its accuracy is high and has high robustness,which provides accurate parameters for the control traffic light of intelligent transportation system.