基于网络监控视频流的车流量实时估计方法  

The Real-time Estimation Method of Vehicle Flow Based on Network Monitoring Video Streams

作　　者：程瑶[1] 卜方玲[1] 严宏海 

机构地区：[1]武汉大学电子信息学院,武汉430072

出　　处：《科学技术与工程》2015年第28期32-38,共7页Science Technology and Engineering

基　　金：国家863项目(2013AA122301)资助

摘　　要：智能交通系统(intelligent transportation system,ITS)是未来交通系统发展方向,其中车流量统计属于智能交通研究的主要领域。提出一种基于视频的车流量实时估计方法,利用redis内存数据库高并发、读写速度快的优势,从连续不断的网络视频流提取帧图像,依次存储在内存数据库的队列中,起到一个缓存实时数据流的作用。在运动车辆检测过程中,利用GMM训练背景模型获得背景帧序列改进Vi Be算法的初始化过程,且在开始检测的过程中加快背景更新速度,达到快速消除原始检测中"鬼影"的目的。3个场景的实验表明:提出的方法能快速消除"鬼影",提高了实时车流量估计的准确性。The development direction of future transportation system is ITS（ intelligent transportation system）,and statistics of the traffic flow belong to one of the prime intelligent transportation＇s research field. Real-time traffic estimation method based on video streams is proposed,making use of redis in-memory database which has high concurrent and great read-write speed,in successive network video streaming extracting frame image and saving in a queue of the in-memory database orderly to cache real-time data stream. During the moving vehicle detection process,by improving Vi Be algorithm＇s initialization procedure,the research is based on gaining background frames sequence under GMM training background model,and accelerating the update speed of the backgrounds when the detection begins which is for the purpose of eliminating original detection＇s ghost objects quickly. The experiment of three scenes shows that the proposed method can quickly remove ＂ghosts＂,and the accuracy of the real-time estimation for the traffic flow is improved.

关 键 词：车流量估计 ViBe算法 鬼影消除 帧队列 网络视频流 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]