摘要
针对实际应用中车道标志线和车辆阴影等无关信息严重影响排队长度检测精度的问题,提出了一种基于图像处理的车辆排队长度鲁棒检测算法;首先将背景差法和帧间差法相结合进行停车状态识别;在此基础上,综合利用梯度差分和颜色差分获取车辆完整信息,从而为采用伸缩窗进行排队长度检测提供了更加鲁棒的信息;实验结果表明算法检测准确,满足实时性的要求。
This paper proposes a novel robust detection algorithm of vehicle queue based on image processing. The paper firstly combines the two methods of background subtraction and frame differencing to recognize vehicle movement states. On this basis, the paper utilizes gradient differences and color differences for complete information of vehicles, which provides more robust information for using the flexible window method to detect queue length. Experimental results show that the algorithm satisfies real--time requirements and show good performance
出处
《计算机测量与控制》
CSCD
北大核心
2011年第8期1810-1813,共4页
Computer Measurement &Control
基金
国家自然科学基金资助项目(60774037)
国家自然科学基金青年基金资助项目(60904069)
教育部博士点新教师基金(20091103120008)
关键词
图像处理
停车状态检测
排队长度检测
伸缩窗
image processing
vehicle movement states
vehicle queue
flexible window
