【作者】 高军；

【导师】 张旗； 王兵；

【作者基本信息】 大连海事大学 ， 控制工程， 2014， 硕士

【摘要】 随着世界经济的快速增长以及各国综合国力的不断增强,电力、交通、银行、安检以及军事设施等领域对安防系统的需求与日俱增且要求也越来越高,视频监控在生产生活的各个方面得到了十分广泛的应用。为了能实时分析、跟踪、判别监控对象,并在异常事件发生时实时提示或报警,为相关部门作出及时决策和正确行动提供支持,视频监控的“智能化”就显得尤为重要。伴随着计算机性能的提高以及机器视觉学科的发展,基于视频分析的智能监控系统应运而生。本课题基于视频分析的处理流程,在分析了智能监控系统中的遗留及遗失物检测功能的需求基础上,依据目标检测、目标跟踪、目标分析这几个视频分析中常用的处理步骤分别进行了相应的算法分析与研究。针对由于光照因素带入的阴影问题,通过采用基于混合高斯的背景模型和阴影模型相结合的方法去除阴影,准确提取前景目标；针对实际场景中经常发生的因遮挡导致漏检的情况,通过基于区域匹配跟踪的方法判断处理遮挡问题；针对行人在场景中长期驻留被误检为遗留物或者遗失物的情况,通过人体判别的方法去除静止人体的干扰；最后利用遗失物和遗留物在特定区域内时间域和空间域上的特点实现了二者的有效区分。本课题同时设计并编程实现了基于视频分析的遗留及遗失物检测的软件系统。在对系统进行需求分析的基础上,采用VC++结合OpenCV的开发方式,设计了系统界面,实现了结果显示、参数调整、控制操作等功能。更多还原

【Abstract】 With the rapid growth of world economy and the strengthening of comprehensive national strength, demands for security system of electric power, transportation, bank, security and military facilities and other areas grow with each passing day. At the same time, the requirement become higher and higher. In a word, video surveillance has been widely used in various aspects of life. In order to analyse, track, discriminate monitoring objects in real time, immediately prompt or alarm when an abnormal event occurs, and provide support to related departments to make timely decisions and right action, the intelligent of video surveillance is particularly important. With the development of computer performance and the improvement of the machine vision research, the intelligent monitoring system based on video analysis emerges at the right moment.On the basis that after analysing the demand of abandoned and stolen object detection function in intelligent monitoring system, this topic has carried on the analysis and research of the corresponding algorithm about the common processing steps of video analysis, including target detection, target tracking, and target analysis.In view of the shadow problems caused by light factors, through the use of background model and shadow model based on mixture Gaussian, the topic removes the shadow successfully and extracts the foreground object accurately; In view of the situation that occasional checking failures often occur in actual scene due to occlusion, through the tracking method based on region matching, the topic deals with the problem finally; In view of the situation that pedestrians are mistakenly identified as abandoned objects or stolen objects due to they stay for a long time in the scene, through people detection, the topic removes the interfere to detection; At last, through the characteristics of abandoned object and stolen object in time domain and space domain of specific areas, the topic distinguishes each other effectively.At the same time, this topic designs and realizes the soft system of abandoned and stolen detection based on video analysis by programming. On the basis of demand analysis of system, this topic designs the system interface by using the development way that VC++combined with OpenCV and realizes the functions of result show, parameter adjustment and control operation.更多还原