【作者】 王宁；

【导师】 游志胜；

【作者基本信息】 四川大学 ， 计算机应用技术， 2005， 硕士

【摘要】 为了解决地面交通快速发展所引发的各种问题,智能交通系统(ITS:Intelligent Transportation System)的研究被提到了重要位置。车辆跟踪与交通事件检测系统作为ITS 的重要组成部分,成为许多国家的研究热点。本文针对ITS 领域中的关键技术,研究了基于视频的车辆跟踪与交通事件检测系统中的相关问题。本文详细描述了基于车辆跟踪结果的交通事件检测方法,分析并研究了车辆逆行、超速、异常停车等交通事件,提出相应的具体检测方法并给出实验结果。论文还就交通事件检测系统给出了系统设计,分析了其中几个值得注意的功能模块,并给出了其框架结构和实现。本文分五个部分进行论述: 1)概述了交通事件检测系统的应用背景与相关技术。对交通事件检测相关技术进行描述和比较。2) 在车辆检测方法中,采用一种改进的基于多样性采样机制的非参数估计背景模型。提取具有较高频度和多样性的灰度新样本集,建立背景核密度估计模型,并利用背景帧差值检测运动车辆。并对车辆检测中阴影问题,车辆粘连、遮挡问题做出详细分析与讨论,并提出了初步的解决方法。通过度量相除后图像的平坦度来检测车辆阴影,修正初步的车辆检测结果。并对车辆外接矩形的求取给出具体的分裂合并规则。3) 基于视频的车辆的跟踪算法。基于车辆检测结果,对一种基于区域匹更多还原

【Abstract】 In order to solve various problems caused by the rapid growing of surface traffic, the research of Intelligent Transportation System comes to an issue of consequence. As an important part of ITS, vehicle tracking and traffic incidents detection becomes a research focus all over the world. According to the key techniques of ITS, some algorithms and techniques for vehicle tracking and traffic incidents detection have been analyzed and improved. Several methods based on vehicle tracking, which are used to detect traffic incidents such as converse and over-speed, have been described in detail in this thesis. What’s more, a vehicle tracking and traffic incidents detection system has been designed for application. The main ideas of this thesis can be described as follow: 1) The first chapter simply introduces the research background of intelligent transportation system, especially techniques in the field of vehicle tracking and traffic incidents detection. 2) For vehicle detection, an algorithm using nonparametric kernel estimation has been analyzed and improved to build the background model and to detect moving objects by background subtraction. By means of measuring the flatness degree of the result of image division,shadows have been eliminated. Therefore, a measurement algorithm which features individual vehicle tracking using rectangular model has been developed to solve problems caused by vehicles overlapping. 3) For vehicle tracking,a method using region matching has been improved and realized. Another algorithm, which is based on improved snake model, has been analyzed and developed to segment moving vehicles and track them. Furthermore, choosing reasonable energy function and parameters has been discussed. 4) Several methods are provided to solve some classical problems about traffic incidents detection such as converse and over-speed. Especially, methods of vehicle speed detection and camera calibration have been analyzed and described in detail. 5) A video vehicle tracking and traffic incidents detection system has been analyzed and structured and main modules of this system are discussed. Design and implementation of this system are described in this chapter. To sum up, according to some difficult points in video vehicle tracking and traffic incidents detection system,some algorithms and methods are proposed, improved to solve those problems. The results of our experiments show that the methods and algorithms proposed in this thesis are adaptable and effective.更多还原