【作者】 宋磊；

【导师】 沈兰荪； 黄祥林；

【作者基本信息】 北京工业大学 ， 电路与系统， 2003， 硕士

【摘要】 视频监控在国民经济中发挥着越来越重要的作用，如何对监控系统中的图像序列进行实用化的智能场景检测分析和智能化管理，实现信息实时有效的获取和控制，是视频监控网络化、智能化的核心技术。 本论文在对视频监控技术的构成与发展进行研究分析的基础上，针对视频监控中涉及的异常检测、异常分析、智能查询和流式传输等若干关键技术进行了深入研究，并建立了视频监控系统的演示平台。 论文根据实际监控工程的需要，首先分析了安全监控中的异常情况，设计了静止物体缓慢变化时的监控算法、光线缓变情况下背景的更新策略、背景变化情况下的检测方法、光线缓慢变化情况下的检测以及消除噪声和干扰的算法等多种具有不同应用需求的算法，同时提出快速异常检测算法的评价公式。在对检测出的异常进行时间、坐标、大小、运动轨迹等分析和描述后，论文设计了基于异常图像内容描述的查询系统，建立了异常图像检索数据库，通过在压缩域提取图像特征节约了系统资源的开销，设计的“粗检+细检+分类”的三层细化检索过程提高了检索结果的准确性，友好的人机界面提供便利的异常查询。为满足视频监控中远程实时传输、处理和回放的需要，论文实现了一种局域网视频流传输技术，主要采用了Csocket套接字技术、DirectShow技术、多线程应用与同步处理技术。 通过对视频监控中图像序列的自动异常检测研究，建立基于内容的智能化查询系统，设计了实时流畅的视频流远程传输等。论文对改善监控系统性能、提升监控系统功能，提高视频监控系统的智能化水平和网络化程度作了有益的尝试。更多还原

【Abstract】 Video surveillance is playing an increasingly important role in the national economy. How to perform intelligent scene detection and to realize real-time & effective management of information are the core technologies of video surveillance.On the basis of analysis the structure and development of video surveillance technology, this paper researches on several key technologies involved in video surveillance, such as detection & analysis of anomaly, intelligent query, streaming transfer, etc., and complete the demonstration platform of video surveillance system.According to the practical requirements of engineering, this paper designs various algorithms, such as those for monitoring objects experiencing a slow change, updating backgrounds while rays are changing slowly, detection under changing background, elimination noise. This paper also advances the formula for evaluating the validity of these algorithms. After providing the analysis of the time, coordinates, size, track and so on of these anomalies, this paper designs an query system based on content of anomalous image, which reduces spending on system resource by means of extracting characteristics of image in Compressed-Domain, improves the accuracy of detection results by the 3-tier detection process designed, i.e. "coarse detection + scrutinized detection + classification", and provide a friendly interface. For the purpose of satisfying the requirements of real-time transfer, processing and play back in the video surveillance, this paper implement a kind of intranet streaming video transfer technology, including Csocket and DirectShow technology, the multithreading application and synchronized processing technology as well.Through researches on several key technologies involved in video surveillance, this paper carries out a helpful attempt to enhance functions of monitoring system and improve intelligent level of video surveillance system.更多还原