本文为加拿大滑铁卢大学（作者：KURTIS NORMAN MCBRIDE）的硕士论文，共80页。

在遮挡与杂乱的环境中，车辆跟踪是一个活跃的研究领域，这种环境中跟踪车辆的问题带来了各种各样的挑战。这些挑战包括车辆跟踪初始化、跟踪未知数量的目标以及现实世界中照明的变化、场景条件和摄像头本身的特性等。场景杂乱和目标遮挡带来了更多其它的挑战，本文提出了一种随机框架，允许从一系列图像中识别车辆的轨迹。本文的工作重点是识别交通场景中存在的车辆轨迹，即发生交叉的车辆运动；该框架结合了背景减法和基于运动历史的分割方法。采用蒙特卡罗-马尔可夫链数据关联（MCMCDA）方法解决了车辆的跟踪问题，该方法包括一个新的概念，即在MCMC评分函数中包含离散、独立区域的概念。研究结果表明，该框架能够跟踪多个车辆相互遮挡、被前景场景对象遮挡的车辆。

Vehicle tracking in environments containingocclusion and clutter is an active research area. The problem of trackingvehicles through such environments presents a variety of challenges. Thesechallenges include vehicle track initialization, tracking an unknown number oftargets and the variations in real-world lighting, scene conditions and cameravantage. Scene clutter and target occlusion present additional challenges. Astochastic framework is proposed which allows for vehicles tracks to beidentified from a sequence of images. The work focuses on the identification ofvehicle tracks present in transportation scenes, namely, vehicle movements atintersections. The framework combines background subtraction and motion historybased approaches to deal with the segmentation problem. The tracking problem issolved using a Monte Carlo Markov Chain Data Association (MCMCDA) method. The methodincludes a novel concept of including the notion of discrete, independentregions in the MCMC scoring function. Results are presented which show that theframework is capable of tracking vehicles in scenes containing multiplevehicles that occlude one another, and that are occluded by foreground sceneobjects.

1 引言
2 项目背景
3 目标跟踪算法
4 实验结果
5 结论

下载英文原文地址：

http://page2.dfpan.com/fs/9lecfjc282d1c289166/

更多精彩文章请关注微信号：