国防科学技术大学硕士学位论文基于移动网络的智能视频监控系统的设计姓名：高成申请学位级别：硕士专业：信息与通信工程指导教师：蔡宣平2010-11国防科学技术大学研究生院硕士学位论文 第 i 页 摘 要 视频监控和移动网络技术的发展与融合推动了移动视频监控技术的快速发展，随着商用 3G 网络的正式运营以及智能移动终端性能的增强，随时随地对远程场景的高速视频监控得以实现。视频中的目标识别理解技术的加入使得监控的实用性得到进一步拓展，移动智能视频监控具有广泛的应用前景。 针对网络监控的具体应用背景，本文提出了一种基于移动网络的智能视频监控系统的设计方案，主要包括运动目标检测、智能告警信息发送以及手机端视频查看等功能模块，论文最后对方案的可行性与稳健性进行了分析。 根据智能监控的应用要求，本文提出了一种基于移动网卡的运动目标智能告警模块的设计方法。针对监控场景中可能出现的运动目标，在分析了目前常用的背景差、帧差法等运动目标检测算法的基础上，采用基于混合高斯模型的背景差算法作为检测标准。结合多线程技术和移动网卡设计出短信智能告警模块，将检测前景的异物点数作为危险行为告警的触发标准，当判定有运动目标后即利用告警模块发送告警信息至手机监控端。 本文的手机应用开发以 Java ME 为平台，对手机客户端用户界面、网络通信以及视频解码与回显等功能进行了分析与设计。通过对比研究选择了适用于远程视频监控的编码方式以及传输协议，实现了手机客户端的实时监控功能。 论文最后给出了一个基于 OpenCV 与 Java ME 开发平台的移动视频监控系统的实现框架，实验结果证明了本文设计方案的有效性。 主题词：智能监控，运动目标检测，混合高斯模型，OpenCV，短信告警，手机安防，Java ME 国防科学技术大学研究生院硕士学位论文 第 ii 页 ABSTRACT The development and integration of video surveillance and mobile network is promoting the naissance and development of mobile video surveillance. With the officially commercial use of 3G network and the rapid development of the mobile intelligent terminal, the high-speed video surveillance towards the remote scene at anytime and anywhere can be achieved. The practicability of surveillance can be widely expanded while using the technique of detecting and understanding moving objects in video, the mobile intelligent video surveillance system will have a wide range of application in times to come. This paper presents a method to design mobile intelligent video surveillance system based on mobile network aiming at the development trend of network surveillance. The main work is the research of the moving object detection algorithm, intelligent alarm message transmission and locale examining with the mobile telephone, then the feasibility and stability of the project is analyzed. According to the application of intelligent video surveillance, this paper presents an intelligent alarm information transmission method based on wireless network module. Aiming at the appearance possibility of moving objects, while giving a detailed analysis for the existing moving object detection algorithms which are widely used, such as background subtraction and frame subtraction, Mixed Gaussian Model is determined as detecting standard. This paper schemes out an intelligent SMS alarm module to detect moving objects using multithreading and mobile network card. The amount of the foreground image luminosity is determined as trigger standards to dangerous behaviors alarm, which can let the server transmit effective alarms after detecting the mobile objects. In this paper, the development platform of mobile applications is Java ME. The main work includes: the analysis and design of user interface, network communications, video decoding and video replay etc. of the mobile phone client. We selecte decoding standard and network transmission protocol suited for remote video surveillance after compareing with other methods, then we realize the realtime surveillance function of the the mobile phone client. Finally, OpenCV and Java ME are used to realize the system discussed above, actual experimental results validated the efficiency of the proposed system. Key Words：Intelligent Surveillance，Moving Object Detection，Mixed Gaussian Model，OpenCV，SMS Alarm，Mobile Security，Java ME 国防科学技术大学研究生院硕士学位论文 第 III 页 表 目 录 表 3.1 室内某实验室测试结果表. 36 表 3.2 室外某公路测试结果表. 37 国防科学技术大学研究生院硕士学位论文 第 IV 页 图 目 录 图 2.1 面向连接的 Socket 系统调用. 12 图 2.2 面向无连接的 Socket 系统调用. 13 图 2.3 RTP 包组成结构. 15 图 2.4 RTP/RTCP 网络层次关系图. 15 图 2.5 监控系统功能模块. 16 图 2.6 监控系统框图. 17 图 2.7 会话信息流. 18 图 2.8 报警信息流. 18 图 2.9 视频信息流. 19 图 2.10 服务器运行流程图. 19 图 2.11 服务器端通信流程图. 20 图 2.12 视频数据发送模块流程图. 21 图 2.13 手机客户端数据流图. 22 图 2.14 客户端运行流程图.