《基于视频的实时运动目标检测与跟踪系统研究硕士学位论文》由会员分享，可在线阅读，更多相关《基于视频的实时运动目标检测与跟踪系统研究硕士学位论文（95页珍藏版）》请在金锄头文库上搜索。
1、 基于视频的实时运动目标检测与跟踪系统研究 硕士学位论文基于视频的实时运动目标检测与跟踪系统研究 Classified Index: UDC: 密级： Xihua UniversityMaster Degree Dissertation Research of Real-time Moving Object Detection and Tracking System Based on VideoCandidate : Li YuejingMajor : Signal and Information ProcessingStudent ID: 212009081002007Supervisor:
2、 Prof. Xie WeichengMarch, 2012西华大学学位论文独创性声明作者郑重声明：所呈交的学位论文，是本人在导师的指导下进行研究工作所取得的成果。尽我所知，除文中已经注明引用内容和致谢的地方外，本论文不包含其他个人或集体已经发表的研究成果，也不包含其他已申请学位或其他用途使用过的成果。与我一同工作的同志对本研究所做的贡献均已在论文中做了明确的说明并表示了谢意。若有不实之处，本人愿意承担相关法律责任。学位论文作者签名： 指导教师签名：日期： 日期西华大学学位论文版权使用授权书本学位论文作者完全了解学校有关保留、使用学位论文的规定，在校攻读学位期间论文工作的知识产权属于西华大学，
3、同意学校保留并向国家有关部门或机构送交论文的复印件和电子版，允许论文被查阅和借阅，西华大学可以将本论文的全部或部分内容编入有关数据库进行检索，可以采用影印、缩印或扫描等复印手段保存和汇编本学位论文。(保密的论文在解密后遵守此规定)学位论文作者签名： 指导教师签名：日期： 日期西华大学硕士学位论文摘 要以计算机视觉技术为基础的智能视频监控系统，目前已经广泛应用于人们生产生活的各个方面。运动目标检测与跟踪是智能视频监控系统中的关键技术和主要研究方向，目的是实时检测视频序列帧中出现的运动目标，获得目标参数，对运动目标进行匹配和跟踪，进而获得目标的运动轨迹。本文主要研究了视频运动目标检测与跟踪系统的构
4、成，系统中使用的视频、图像处理技术，运动目标检测及运动目标跟踪算法，并分别在计算机系统及嵌入式系统中实现了运动目标检测与跟踪系统。首先，分析研究了常用的运动目标检测算法及各自的优缺点，提出了一种基于累积差分更新的背景减除法，该算法结合累积差分的概念，对自适应背景建模法作出改进，较好地消除了场景变化以及噪声影响等的干扰，关于目标提取的阀值判断，提出了一种两主峰间差值的灰度直方图阀值分割法。接着，分析比较了连续自适应均值漂移CAMShift算法与Kalman滤波算法，提出了一种基于几何特征的Kalman滤波与目标直方图匹配相结合的运动目标跟踪算法，以目标灰度质心及外接矩形框长宽作为目标几何特征参数
5、，分别使用两组卡尔曼滤波预测后，根据目标灰度模板相似度进行目标匹配。然后，在VC+6.0集成开发环境下，采用MFC应用程序框架及OpenCV计算机视觉库代码，实现了基于USB摄像头的运动目标检测与跟踪系统，对于USB摄像头或AVI视频文件输入的视频，系统能实时检测出场景中的运动物体并进行跟踪。最后，使用Altera ED2多媒体开发板，以Cyclone系列EP2C35F672C6 FPGA作为核心芯片，设计出基于SOPC的实时运动目标检测与跟踪系统，使用了Quartus、SOPC Builder及NiosIDE等开发工具，完成硬件设计及软件开发，经仿真及调试该系统实现部分运行。关键词：智能视频
6、监控；目标检测；目标跟踪；OpenCV；SOPC I AbstractIntelligent Video Surveillance System which based on Computer Vision technology has already been widely applied to various aspects of humans production and life at present. Moving object detection and tacking is the key technology and primary research direction of
7、Intelligent Video Surveillance System. In order to real-time detecting moving object emerged from video sequence frames, obtaining object parameters, matching and tracking moving object, then the movement locus of object is acquired. In this paper major researches include the structure of moving obj
8、ect detection and tracking system based on video, the technology of video and image processing used in system, the algorithm of moving object detection and tracking, and the implementation of system based on computer and embedded technology. Firstly, conventional algorithms of moving object detectio
9、n and tracking including each advantage and defect are analyzed and researched, and a kind background subtraction algorithm based on accumulated difference update is proposed. The algorithm incorporated accumulated difference notion improvements self-adaptive background modeling algorithm to elimina
10、te noise interference. A kind partition method of gray histogram threshold is proposed.Then, CAMShift algorithm and Kalman filter algorithm are analyzed and compared, and a kind moving object tracking algorithm which combines Kalman filter based on geometric characteristic with object histogram matc
11、hing is proposed. The center of mass of objects grayscale and the length and width of circumscribed rectangle frame are used as the geometric characteristic parameter of object. After using respectively two groups of Kalman filter to calculate, matching object according to similarity of template of
12、objects grayscale.And then, the moving object detection and tracking system based on USB camera is implemented by VC+ 6.0 Integrated Development Environment. The MFC application program wizard and Intel OpenCV code are utilized to develop system. According to inputted video from either USB camera or
13、 AVI files, the system can detect and track object in scene. Finally, the real-time moving object detection and tracking system based on SOPC is designed. The Altera DE2 board which contains Cycloneseries EP2C35F672C6 FPGA chip is used as development platform. The Developer Kits include Quartus, SOP
14、C Builder and NiosIDE etc. The hardware design and software programming are accomplished. The embedded system is able to partly operate after simulating and debugging.Key Words：Intelligent Video Surveillance; Object Detection; Object Tracking; Open CV; SOPC III 目 录摘 要IAbstractII1 绪论11.1 研究背景及意义11.2 国内外研究现状及发展趋势21.2.1 国内外研究现状21.2.2 发展趋势41.3 主要研究内容及本文结构62 系统结构与关键技术82.1 系统结构82.2 图像处理技术92.2.1 图像获取92.2.2 视频信号制式及图像色彩模型102.2.3 图像滤波132.2.4 形态学处理162.2.5 连通分量分析183 运动目标检测算法研究203.1 常用运动目标检测算法分析203.1.1 帧间差分法203.1.2 背景减除法223.1.3 光流法233.2 基于累积差分更新的背景减除法233.2.1 背景模型建立与更新243.2.2 运动目标提取273.3 检测算
