【信息技术】【2004.08】智能视频监控中的运动目标检测、跟踪与分类

本文为土耳其比尔肯大学（作者：Yi˘githan Dedeo˘glu）的硕士论文，共100页。

长期以来，视频监控一直用于监控安全敏感地区，如银行、百货公司、高速公路、拥挤的公共场所和边界地区。计算能力、大容量存储设备的发展以及高速网络基础设施的进步为更便宜的多传感器视频监控系统铺平了道路。传统上，监控视频输出由操作人员在线处理，并保存到磁带上以便取证使用。普通监控系统中照相机数量的增加使操作人员和存储设备都超载了大量的数据，从而无法确保对敏感区域的长期监测。为了滤除由照相机阵列产生的冗余信息，增加对取证事件的响应时间，利用“智能”视频监控系统辅助操作人员识别视频中的重要事件已经成为一个关键要求。视频监控系统“智能化”的实现需要快速、可靠和鲁棒的运动目标检测、分类、跟踪和活动分析算法。

本文提出了一种具有实时运动目标检测、分类和跟踪能力的智能视觉监控系统，该系统对静止相机的彩色和灰度视频图像进行处理。它可以处理室内和室外环境中的物体检测，并且可以在变化的照明条件下工作。分类算法利用所检测目标的形状和时间特征进行跟踪，能够将目标成功分为单人、人群和车辆等预定义类别。该系统还能够可靠地检测各种场景中的自然现象：火灾。即使存在偶尔的完全遮挡，所提出的跟踪算法也能成功地跟踪视频目标。除此之外，还能够满足鲁棒智能视频监控系统的一些重要需求，例如去除阴影、检测突然的照明变化以及区遗留/移除的目标。

Video surveillance has long been in use tomonitor security sensitive areas such as banks, department stores, highways,crowded public places and borders. The advance in computing power, availabilityof large-capacity storage devices and high speed network infrastructure pavedthe way for cheaper, multi sensor video surveillance systems. Traditionally,the video outputs are processed online by human operators and are usually savedto tapes for later use only after a forensic event. The increase in the numberof cameras in ordinary surveillance systems overloaded both the human operatorsand the storage devices with high volumes of data and made it infeasible toensure proper monitoring of sensitive areas for long times. In order to filterout redundant information generated by an array of cameras, and increase theresponse time to forensic events, assisting the human operators withidentification of important events in video by the use of “smart” videosurveillance systems has become a critical requirement. The making of videosurveillance systems “smart” requires fast, reliable and robust algorithms formoving object detection, classification, tracking and activity analysis. Inthis thesis, a smart visual surveillance system with real-time moving objectdetection, classification and tracking capabilities is presented. The systemoperates on both color and gray scale video imagery from a stationary camera.It can handle object detection in indoor and outdoor environments and underchanging illumination conditions. The classification algorithm makes use of theshape of the detected objects and temporal tracking results to successfullycategorize objects into pre-defined classes like human, human group andvehicle. The system is also able to detect the natural phenomenon fire invarious scenes reliably. The proposed tracking algorithm successfully tracksvideo objects even in full occlusion cases. In addition to these, someimportant needs of a robust smart video surveillance system such as removingshadows, detecting sudden illumination changes and distinguishing left/removedobjects are met.

1 引言
2 智能视频监控调研
3 目标检测与跟踪
4 目标分类
5 火灾检测
6 实验结果
7 结论与未来工作展望

下载英文原文地址：

http://page5.dfpan.com/fs/1lcj9221e291a651b45/

更多精彩文章请关注微信号：