基于视频分析的道路交通状态检测研究

发布时间：2018-09-03 07:09

【摘要】：随着社会与经济的迅速发展,城市规模日益扩大,人口与道路车辆数目迅速增加,导致交通事故、交通拥堵等问题越来越严重,环境污染及资源浪费等问题也随之而来,对城市发展及人们的生活质量产生了极大的负面影响。因此,道路交通状态检测成为交通视频监控中研究的重要内容之一。本文提出了一种基于视频分析的道路交通状态检测方法。通过对道路监控视频的分析,研究了视频图像中运动车辆检测、交通特征参数提取及交通状态判断等问题。主要工作有如下几方面:(1)对混合高斯背景建模进行改进,提高了背景模型对光线变化的适应性,较好地检测出运动车辆目标,并结合背景差分方法获取车辆运动区域;(2)提出了一种基于边缘特征检测道路车辆密度的方法,避免了对车辆区域的分割,降低了参数提取的难度;(3)对交通状态进行道路交通流宏观描述,将块匹配算法用于估计交通流速度。在此基础上,采用K-均值聚类算法设计两层分类器,对三种交通状态(畅通、缓行与拥堵)进行分类识别。(4)将建议算法集成系统,对贵阳市实际道路交通监控视频进行了测试,实验结果表明,建议算法有效、可行,为交通状态识别的后续研究奠定了一定的基础。
[Abstract]:With the rapid development of society and economy, the city scale is expanding day by day, the population and the number of road vehicles are increasing rapidly, which leads to more and more serious problems, such as traffic accidents, traffic congestion, environmental pollution and waste of resources and so on. It has a great negative impact on urban development and people's quality of life. Therefore, road traffic state detection has become one of the important contents in traffic video surveillance. In this paper, a method of road traffic state detection based on video analysis is proposed. Based on the analysis of road surveillance video, the problems of moving vehicle detection, traffic feature parameter extraction and traffic state judgment in video image are studied. The main works are as follows: (1) the background modeling of mixed Gao Si is improved to improve the adaptability of the background model to light change and to detect the moving vehicle target. Combined with background difference method, the moving area of vehicle is obtained. (2) A method based on edge feature is proposed to detect vehicle density, which avoids the segmentation of vehicle area. The difficulty of parameter extraction is reduced. (3) the traffic state is described macroscopically and the block matching algorithm is used to estimate the speed of traffic flow. On this basis, a two-layer classifier is designed by using the K-means clustering algorithm to classify and identify the three traffic states (smooth, slow and congested). (4) the proposed algorithm is integrated into the system, and the actual road traffic surveillance video in Guiyang is tested. The experimental results show that the proposed algorithm is effective and feasible, which lays a foundation for further research on traffic state recognition.
【学位授予单位】：贵州民族大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】