基于内容的监控视频检索算法研究

发布时间：2018-04-18 12:07

本文选题：监控视频检索 + 镜头分割　；参考：《山西大学》2014年硕士论文

【摘要】：视频监控是通过摄像头来获取一定区域的视频图像信息,以实现对进入该区域范围内的目标及其行为动态进行监督的目的。目前,视频监控已经日益广泛地应用在国计民生的多个领域中,如安防、交通、军事等。视频监控系统产生了海量的视频文件,由于视频文件数据量大、结构复杂、表现形式多样化,人们使用传统的基于文本标记的浏览和检索方式来访问监控视频,无论在时耗和精度上都很难满足实际工作的需求。针对这一问题,本文对一种高效的浏览、检索监控视频的算法即基于内容的监控视频检索算法展开研究。本文在研究基于内容的视频检索算法基本理论的基础上,结合视频监控场景的特点和人们的检索需求,重点对镜头分割、关键帧提取、关键帧检索匹配等视频检索中的关键技术进行研究。主要的研究内容如下文所述：(1)研究分析基于内容的视频检索算法的应用现状和发展前景,回顾、展望该领域的国内外发展动态,对其基础理论知识和一些常用的检索算法进行研究分析。(2)结合视频监控场景的特点和实际需要,研究提出了一种基于灰度变化检测的镜头分割算法。通过设定虚拟检测线,统计计算虚拟检测线路径上灰度变化来确定镜头的开始；计算目标前景总灰度值,当其减小到一定值时,镜头结束。以此获得一个完整的镜头。(3)研究关键帧提取方法,在提取监控视频镜头关键帧时,首先合理的选取第一个关键帧,再统计边缘方向直方图、计算帧间差来更新、获取其余的关键帧。(4)研究关键帧的检索匹配算法,研究提出一种基于边缘方向直方图相关性匹配的图像检索算法。对图像进行去噪、提取边缘后,计算获取边缘方向直方图,等级化排列直方图构成特征向量。再使用斯皮尔曼等级相关公式计算图像特征向量间的相关系数作为衡量图像间相似性的指标。通过实验对算法的有效性、可靠性进行验证。(5)鉴于人们常关注监控视频中目标的颜色、形状信息,研究了一种综合使用颜色特征和形状特征的关键帧匹配算法并通过实验验证了算法的性能。最后,结合网站开发相关技术和本文研究的算法,研究开发一个在线的监控视频检索系统,用户可以远程登录系统对视频进行检索。
[Abstract]:Video surveillance is to obtain the video image information of a certain area through the camera, in order to achieve the purpose of monitoring the target and its behavior in the region.At present, video surveillance has been widely used in many fields of national economy and people's livelihood, such as security, traffic, military and so on.Video surveillance system has produced a large number of video files. Because of the large amount of data, complex structure and diverse forms of expression, people use the traditional browsing and retrieval methods based on text markup to access the monitored video.It is difficult to meet the requirements of practical work in terms of time consumption and accuracy.In order to solve this problem, this paper studies an efficient browsing and retrieval algorithm for surveillance video, that is, content-based video retrieval algorithm.Based on the research of the basic theory of content-based video retrieval algorithm and the characteristics of video surveillance scene and people's retrieval requirements, this paper focuses on shot segmentation and key frame extraction.Key techniques in video retrieval such as key frame retrieval and matching are studied.The main research contents are as follows: (1) Research and analysis of the application status and development prospects of content-based video retrieval algorithms, review, and prospects for the development of this field at home and abroad.Based on the basic theoretical knowledge and some commonly used retrieval algorithms, a shot segmentation algorithm based on gray change detection is proposed, which is based on the characteristics and practical needs of video surveillance scene.By setting the virtual detection line, the grayscale change on the path of the virtual detection line is statistically calculated to determine the start of the shot, and the total gray value of the target foreground is calculated, when it is reduced to a certain value, the shot ends.In order to obtain a complete shot. 3) the key frame extraction method is studied. When extracting the key frame of surveillance video shot, the first key frame is selected reasonably, then the edge direction histogram is counted and the difference between frames is calculated to update.The key frame matching algorithm is studied, and an image retrieval algorithm based on edge direction histogram correlation matching is proposed.The image is de-noised, the edge is extracted, the edge direction histogram is obtained, and the hierarchical histogram is arranged to form the feature vector.Then the correlation coefficient between the image feature vectors is calculated by using the Spelman rank correlation formula as an index to measure the similarity between images.The validity and reliability of the algorithm are verified by experiments. (5) in view of the fact that people often pay attention to the color and shape information of the target in the surveillance video,A key frame matching algorithm which combines color features with shape features is studied and the performance of the algorithm is verified by experiments.Finally, an online surveillance video retrieval system is developed by combining the related technologies of website development and the algorithms studied in this paper. Users can remotely log on to the video retrieval system.
【学位授予单位】：山西大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN948.6

【参考文献】