视频道路交通信息处理关键技术研究

发布时间：2018-03-30 10:43

本文选题：交通视频　切入点：车辆检测　出处：《湖南大学》2014年硕士论文

【摘要】：实时采集道路网络中的道路交通流量、车速、车流密度、车头时距、车辆行程时间、以及占用率等宏观和微观的道路交通信息，可以帮助掌握道路交通流的规律，指导人们进行道路交通管理与控制。道路交通信息是道路交通设施改善、道路交通规划等必要的基础数据。因此，建立道路交通信息采集方法是一个很值得研究的课题。本文的主要研究内容如下：（1）采用统计方法更新视频图像背景，同时为了适应道路交通的复杂性，对不同道路环境下背景更新过程中相关参数进行了讨论和分析，并给出了参考值。在获取道路背景后，利用邻域、多阈值的方法对传统背景差分法来进行改进，克服其在运动目标检测过程中的不足。然后通过融合图像阴影的多种特征，，获得包含视频图像阴影和亮度较低车辆的区域，再构建一组向量消除视频图像阴影的影响。通过详细的实验进一步验证采用改进的背景差分法和本文提出的阴影消除方法对车辆检测结果的可靠性。（2）为消除视频图像中车辆之间相互遮挡的影响，提高道路交通信息检测的准确性，提出了一种基于数学形态学的方法对视频图像中车辆之间相互遮挡的图像进行分割。计算视频图像中的车辆实体面积与最小外接多边形的面积差，以及视频图像中单个连通区域内车辆面积来智能识别遮挡是否存在。存在遮挡时，通过邻域特征搜索到凹包的顶点，然后利用凹包的顶点，作为腐蚀起点的判断依据，再腐蚀掉在距离变换过程中受凹包顶点影响的像素点。在完成对目标车辆遮挡识别和相关处理后，本文融合基于特征匹配的追踪方法和基于灰色系统的预测方法对运动目标追踪，先利用特征匹配追踪方法对运动车辆进行初步追踪，并利用灰色预测进行辅助判断特征匹配结果的可靠性。（3）采用计算机编程将相关理论和技术转化为实践，本文设计并实现了道路交通信息采集的软件。选取了代表近距离拍摄和远距离拍摄的两个交通视频实例对该技术进行检验和分析，获取了交通量、车速、车头时距、交通密度，车辆达到时间等交通信息。
[Abstract]:The real-time collection of macro and micro road traffic information, such as road traffic flow, speed, vehicle density, headway time, vehicle travel time, and occupancy rate, can help to master the laws of road traffic flow. Guiding people to carry out road traffic management and control. Road traffic information is the necessary basic data such as road traffic facilities improvement, road traffic planning and so on. The establishment of road traffic information collection method is a topic worthy of study. The main contents of this paper are as follows:. In order to adapt to the complexity of road traffic, the relevant parameters in the background updating process under different road environments are discussed and analyzed, and the reference values are given. The traditional background difference method is improved by using neighborhood and multi-threshold method to overcome the shortcomings in moving target detection. To obtain an area containing shadow and low brightness vehicles in the video image, Then a set of vectors is constructed to eliminate the shadow effect of video images. The reliability of the improved background difference method and the shadow cancellation method proposed in this paper is further verified by the detailed experiments. In order to eliminate the influence of mutual occlusion between vehicles in video images and improve the accuracy of road traffic information detection, In this paper, a mathematical morphology based method is proposed to segment the images of mutual occlusion between vehicles in video images, and to calculate the difference between the area of vehicles in video images and the minimum external polygon. When there is occlusion, the vertex of concave packet is searched by neighborhood feature, and then the vertex of concave packet is used as the basis for judging the corrosion starting point. Then erode the pixels affected by the concave vertex during the distance transformation. After the target vehicle occlusion recognition and correlation processing are completed, In this paper, the tracking method based on feature matching and the prediction method based on gray system are combined to track the moving target. Firstly, the tracking method of feature matching is used to track the moving vehicle. The reliability of feature matching results is evaluated by grey prediction. Using computer programming to translate relevant theories and technologies into practice, In this paper, the software of road traffic information collection is designed and implemented. Two traffic video examples, which represent close shooting and long distance shooting, are selected to test and analyze the technology, and the traffic volume, speed, headway time distance and traffic density are obtained. Vehicle arrival time and other traffic information.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495

【参考文献】