摘要
目前,高速公路交通拥堵日趋频繁,然而大部分高速公路的管理方式仍然是通过人工查看轮巡监控视频来发现交通拥堵,效率低下。本文提出一种对高速公路轮巡监控视频自动检测交通拥堵的方法。该方法实时检测监控视频场景是否轮巡、切换、变动,自适应提取变动场景的道路边界结构并分类,然后在提取的道路边界结构范围内,计算宏观边缘占有率与加权平均光流速度两个宏观交通状态参数,并构造特征集,根据不同的道路边界类型选择对应支持向量机来实现对交通拥堵的自动检测和判别。实验结果表明,该方法能有效地对轮巡模式工作的高速公路监控视频检测交通拥堵,检测时间不大于30s,检测正确率达91.8%。
Traffic congestion is increasingly serious in expressway, and manual surveillance is still a main way to find traffic congestion, which is very ineffective. In this paper, an automatic traffic congestion detection method for expressway surveillance videos is presented. Firstly, the road boundary structure of surveillance video is extracted adaptively. Secondly, edge occupancy rate and weighted average velocity of optical flow in the range of extracted road are calculated. Finally, the traffic congestion is detected by analyzing the two macro traffic state parameters. The experimental test results show that this method is effective for express- way switching surveillance video, the detection time is less than 30 seconds, and the detection accuracy rate is above 91.8%.
出处
《计算机与现代化》
2013年第11期28-33,共6页
Computer and Modernization
基金
国家自然科学基金资助项目(51078362)
公安部应用创新计划项目(2011YYCXGDST078)
关键词
交通拥堵检测
边缘占有率
宏观光流速度
高速公路
轮巡监控视频
traffic congestion detection
edge occupancy rate
macro optical flow velocity
expressway
switching surveillance video