本文为美国堪萨斯大学（作者：Manjunath Narayana）的硕士论文，共134页。
自动监视系统对于安全领域至关重要。视频监控中运动目标的可靠检测与跟踪是高级智能应用的基础，具有很多开放性问题。我们的工作重点是开发一个处理框架，以检测运动物体并从现实世界的监控视频产生可靠的运动轨迹。在建立了基础系统作为进一步自动跟踪研究的平台之后，我们研究了监控视频中摄像机与场景不同位置(对象深度)对象之间的距离变化问题。基于场景不同位置的目标运动，提出了一种基于反馈的解决方案，用于在静态摄像机视频场景中自适应地找出距离的变化信息。该方案基于我们称之为“邻近因子”的概念，具有对噪声和目标分割问题的鲁棒性。邻近因子也可应用于估计场景不同部分中的空间阈值和对象大小。此外，开发了一种新的贝叶斯算法来跟踪视频中目标的轨迹；贝叶斯方法允许概率跟踪分配，可以作为未来更高层次推理的基础。
Automated surveillance systems are ofcritical importance for the field of security. The task of reliably detectingand tracking moving objects in surveillance video, which forms a basis forhigher level intelligence applications, has many open questions. Our workfocuses on developing a framework to detect moving objects and generatereliable tracks from real-world surveillance video. After setting up a basicsystem that can serve as a platform for further automatic tracking research, wetackle the question of variation in distances between the camera and theobjects in different parts of the scene (object depth) in surveillance videos.A feedback-based solution to automatically learn the distance variation instaticcamera video scenes is implemented, based on object motion in differentparts of the scene. The solution, based on a concept that we call ’VicinityFactor’, is robust to noise and object segmentation problems. The VicinityFactor can also be applied to estimate spatial thresholds and object size indifferent parts of the scene. Further, a new Bayesian algorithm to assigntracks to objects in the video is developed. The Bayesian method allows forprobabilistic track assignment and can be the basis for future higher-levelinference.
1 引言
2 项目背景
3 系统架构及描述
4 基本算法的缺陷
5 改进算法的处理结果
6 场景深度变化：邻近因子的引入
7 引入邻近隐私的处理结果
8 贝叶斯跟踪算法
9 贝叶斯跟踪的处理结果
10 结论
11 未来研究工作展望
下载英文原文地址：
http://page2.dfpan.com/fs/alcjd221e291e630502/
更多精彩文章请关注微信号：
