【作者】 陈潇君；

【导师】 鞠时光；

【作者基本信息】 江苏大学 ， 计算机应用技术， 2008， 硕士

【摘要】 伴随着网络和数字视频技术的飞速发展,监控技术正向着智能化、网络化方向不断前进。智能视频监控是计算机视觉领域一个新兴的应用方向,智能视频监控是在不需要人为干预情况下,利用计算机视觉和视频分析的方法对摄像机拍录的图像序列进行自动分析,实现对动态场景中目标的定位、识别和跟踪,并在此基础上分析和判断目标的行为。在完成日常管理的同时,对异常情况及时作出反应。从功能上,视频监控可用于安全防范、信息获取和指挥调度等方面。本文围绕着智能视频监控的一些关键技术,包括背景分析、对象提取、对象识别、对象描述、对象跟踪和对象行为分析。以海关码头智能监管系统中的事件检测为应用背景,对运动目标检测、提取和跟踪进行了深入的分析研究,并在此基础上完成了海关码头智能监管系统CDIIS(Custom Dock Intelligent Inspecting System)的设计与开发。本文的主要工作包括:1.提出了一种改进的混合高斯背景模型,并引入平滑方程,用于摄像机静止状态下的运动目标检测,对序列图像进行背景建模,将多个运动目标从背景图像中分离出来。该算法可以有效解决动态环境中捕捉的图像受到天气的变化、光照条件的变化、摄像机微小的抖动等影响。2.提出一种新的基于轮廓线的轮廓矩不变量特征的方法,用于表示运动目标特征。运动目标提取分别采用了外接多边形方法得到运动目标的初始大致轮廓,和Snake算法计算出运动目标的详细轮廓信息。针对CDIIS中海关码头监控中的船体对象的形状特性,提出了轮廓不变量特征方法。试验证明,该方法同传统方法相比,计算时间短,并可以精确的描述出运动物体。3.提出了一种快速CamShift算法,并将其与活动轮廓模型相结合,用于运动目标跟踪,该方法有效解决了对遮掩物体跟踪的问题。在处理过程中引入Epanechikov核函数平滑色彩概率分布图,并证明了其收敛性。实验结果证明在直方图背投影中引入核函数,能够得到平滑的输入信息;快速CamShift算法也具有较好的实时性和较强的抗干扰能力。整个系统在镇江海关下属多个码头运行的结果表明,本文所采用的一整套方法能够达到理想的效果。更多还原

【Abstract】 With the rapid development of network and digital video technology,the monitoring technology is toward intelligent,network-based forward.Intelligent video surveillance is an emerging field of computer vision application.It is no longer required human intervention, with using computer vision and video analysis for automatic analyzing the image sequence, identifying the location and tracking the trajectory of moving target.Meanwhile,the intelligent video surveillance systems could analysis and judge the behavior of moving targets.In the day-to-day management,System could responds of anomalies in a timely manner.From its function,the video surveillance can be used for security,access to information and control,and so on.In this paper,we discuss a number of key technologies for intelligent video surveillance, including background modeling,the object extraction,object recognition,the object description,the object tracking,and the behavior analysis of the object.With the background of intelligent video surveillance for the Customs dock,the paper has an in-depth study on the problem of moving target detection,extraction and tracking,designs and develops the system of Custom Dock Intelligent Inspecting System intelligent(CDIIS).The main work includes:1.An improved Mixture Gaussian model of background modeling has been proposed, and a smoothing equation has been introduced for detecting the moving object,modeling the sequence of images background,and separating the multiple objects from the background scene with the Static camera.The proposed algorithm can resolve the impact of weather and light change,the Camera tiny jitter problems for the captured images in a dynamic environment.2.A new method of the invariant moment features based on the outline,for describing features of the moving target,has been proposed.The extraction of moving target used external polygon method of the initial outline,and the Snake algorithm to calculate a detailed outline of moving objects.According to the features of the objects of the Customs dock,we proposed the new method,which has been proved that it could describe the outline feature of moving objects more accurately and computing more shortly,compared with the traditional method.3.A fast CamShift algorithm with the combination of active Contour model has been proposed,for tracking moving objects,which could solve the cover problem of tracking objects effectively.In the process,we introduce the Epanechikov Kernel function to smooth the color probability distribution,and proved its convergence.The results prove that the Epanechikov Kernel function could smooth the input information.And the fast CamShift algorithm is also better real-time and a strong anti-interference capability.The CDIIS has been running on Zhenjiang Customs dock,the implement results show that a set of methods,we used,to achieve the desired results.更多还原