基于视频序列的人物跟踪与行为分析技术研究

Research on Technologies of Moving Human Object Tracking and Behavior Analysis in Video

作者: 专业:计算机软件与理论 导师:张菁 年度:2010 学位:硕士  院校: 哈尔滨工程大学

Keywords

Moving object detection, Moving human tracking, Behavior analysis, Automatic monitoring, Automatic alarming

        视频中的人物跟踪与行为分析技术是以人为监控目标的监控系统智能化、自动化的基础技术。目前,此类技术在军事、竞技体育等领域已有广泛应用,而在经费支出较少的普通民用领域则发展较慢但发展前景广阔。本文的主要研究内容包括基于视频的人物跟踪与行为分析两方面技术,其中人物行为分析技术的研究是以人物运动跟踪的结果为基础的。对运动的人物目标进行准确定位及跟踪的方法和以提高定位精度为目标的阴影消除算法是本文研究内容的重点。首先对图像的采集、预处理技术进行研究。通过综合比较DirectShow与VFW技术,选择一种高效的数据采集方法。通过合理应用灰度变换、滤波等图像增强技术来解决环境及设备因素造成的图像质量下降问题。其次,对运动目标检测技术进行研究,针对背景差值法易受阴影干扰的缺点给对目标精确定位带来的困难,提出一种通过对阴影位置进行估计来消除其对定位精度影响的算法。对阈值分割方法进行研究,提出能高效确定阈值更新时机的算法来降低频繁更新阈值带来的运算复杂度。此外,通过分析运动目标跟踪算法并结合实际情况,提出一种基于目标形心投影点的目标定位及跟踪方法。该方法可以对目标人物准确定位,较传统采用质心定位的方式有运算简单,定位准确的特点。最后,对人物跟踪得到的数据进行研究,并以此为基础对被监控对象的行为进行分析与识别,进而完成一套人物跟踪与行为识别系统。该系统可以实现对进入预设区域,行走超速,行走路径等进行分析记录,对异常行为进行识别和自动报警。
    The technologies on moving human tracking and behavior analysis are the basis that a system monitors human object to be automated and intellectualized. Some relating technologies have been widely used in the military field and competitive sports field, however, only few ones are used in the common civil monitoring field in which the systems lack fund but have broad prospects.The author pays major attention to two aspects technologies including moving human body tracking and their behaviors analyzing, the latter technology is based on the former one. The most important work among the research in the dissertation is to propose methods to locate the moving human body in the video accurately and eliminate the drop shadow of moving human body in order to improve the quality of location and tracking.Firstly, much research work relating to image capture and image pretreatment will be done in the paper. By the comprehensive comparison of DirectShow and VFW technology, a highly efficient method of data collection is adopted. According to the analyzing aimed at gray-scale transformation, filtering and other image enhancement technologies, the solutions to noise and image blurring problem caused by equipment and environment are determined. Secondly, by the research on moving target detection techniques, against the shortcomings of the background difference method, a method improving location accuracy is put forward to eliminate drop shadow. Benefitting from the analysis of threshold segmentation, an algorithm determining the time of refreshing the threshold is designed to reduce the calculating complexity caused by updating the threshold value frequently. In addition, by analyzing the tracking method on moving target, a tracking algorithm is proposed, which can accurately locate moving human targets. The method’s advantages in contrast to the traditional centroid-oriented positioning methods are simple and precise when used in positioning characteristics. Finally, the research on the object behavior analysis and identification are implemented based on the data obtained from moving human object tracking algorithm. And a behavior analysis and recognition system is accomplished by using technologies proposed in the dissertation, which should identify risk behaviors and automatic alarm if some behaviors have been found, such as entering the preset area, over-speed walking etc.
        

基于视频序列的人物跟踪与行为分析技术研究

摘要5-6
Abstract6
第1章 绪论10-16
    1.1 课题背景及研究意义10-11
    1.2 国内外研究现状及相关技术11-14
        1.2.1 国内外研究现状11-12
        1.2.2 相关技术发展现状12-13
        1.2.3 现存的问题13-14
    1.3 论文的主要工作内容14-15
    1.4 论文的组织结构15-16
第2章 系统框架设计16-24
    2.1 系统设计思想及原则16-18
    2.2 系统的硬件构成及处理顺序18-19
        2.2.1 硬件设备构成18
        2.2.2 系统的总体处理顺序18-19
    2.3 系统的软件处理模块划分19-23
        2.3.1 软件系统组成19
        2.3.2 各功能模块介绍19-21
        2.3.3 主要模块的处理流程21-23
    2.4 系统开发环境23
    2.5 本章小结23-24
第3章 运动目标检测相关技术研究24-44
    3.1 从视频中获取图像所用技术24-26
    3.2 图像预处理技术26-31
        3.2.1 灰度变换26-27
        3.2.2 基于滤波的图像增强方法27-31
    3.3 运动目标检测方法31-34
        3.3.1 背景差值法31-33
        3.3.2 图像差分法33
        3.3.3 光流分割法33
        3.3.4 其他方法33-34
    3.4 阈值分割方法34-37
    3.5 运动目标分类37
    3.6 本章的核心工作37-40
        3.6.1 阴影消除37-39
        3.6.2 阈值的选取时机选择39-40
    3.7 实验与结果分析40-43
        3.7.1 阴影去除实验40-42
        3.7.2 自动阈值选取42-43
    3.8 本章小结43-44
第4章 运动人物定位及跟踪技术研究44-56
    4.1 运动人物定位技术研究44-46
        4.1.1 单摄像机实现目标定位44-45
        4.1.2 多摄像机立体视觉实现目标定位45
        4.1.3 本文提出的目标定位方法45-46
    4.2 人物运动跟踪技术研究46-48
        4.2.1 人物运动跟踪相关问题46-47
        4.2.2 常用运动跟踪算法总结47-48
        4.2.3 本文运动目标跟踪需要解决的问题48
    4.3 本文实现的运动人物目标跟踪方法48-52
        4.3.1 原理及采取的方法49-51
        4.3.2 投影点的具体计算方法51-52
    4.4 实验及结果分析52-55
    4.5 本章小节55-56
第5章 运动人物行为分析技术研究56-67
    5.1 人物行为分析的目的56
    5.2 人物行为分析关注的特征及方法56-62
        5.2.1 目标人物进入特定区域56-60
        5.2.2 目标人物行走路径60-61
        5.2.3 目标人物超速运动61-62
        5.2.4 图像中其它的人物特征62
    5.3 实验及结果分析62-66
        5.3.1 目标人物进入预定区域及超速实验62-66
    5.4 本章小节66-67
结论67-68
参考文献68-72
攻读硕士期间发表的文章和取得的科研成果72-73
致谢73
        下载全文需10


本文地址:

上一篇:柴油机数字式电子调速器的设计与开发研究
下一篇:零码平台业务自动建模的研究与实现

分享到: 分享基于视频序列的人物跟踪与行为分析技术研究到腾讯微博           收藏
评论排行
公告