基于Hadoop架构的分布式视频关键帧提取方法研究

发布时间：2018-01-06 02:22

本文关键词：基于Hadoop架构的分布式视频关键帧提取方法研究　出处：《合肥工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着视频监控技术在各行业领域内被广泛运用,导致了监控视频数据量呈指数增长,人们早已经厌烦了以往根据时间信息手动拖拽进度条的浏览方式,不仅消耗了查询者大量的时间和精力,而且容易遗漏关键目标信息。关键帧提取技术的出现使得这种情况得到了极大的改善,关键帧可以不受时间、音视频同步等问题的影响,并可以提供多种方式进行浏览和导航使用。但是在实际应用关键帧提取技术的过程中,面临的最大困难就是关键帧提取速度太慢：一方面是算法的高复杂性；另一方面是当前使用的是单机模式提取算法。本文在对关键帧提取算法充分调研之后,提出一种新的关键帧提取方法,并在将算法移植到Hadoop云平台的过程中,解决了帧完整性等核心问题,成功将算法改造成了分布式提取方式,极大地提高了关键帧的提取速度。首先,本文指出了关键帧技术对视频检索等实际工程应用的重要性,并介绍了近年来发展如火如荼的云计算技术和其在多媒体领域的应用现状,同时对Hadoop的核心技术和理论进行了充分的学习和调研。其次,本文分析了当前关键帧提取方法存在的问题,提出一种新的自适应的视频关键帧提取方法,该方法能够自适应确定关键帧数目、计算量小并且对内容渐变的视频的处理效果更佳。再次,本文分析了在Hadoop云平台上实现视频关键帧的分布式提取云应用所面临的难题,例如帧的完整性、处理逻辑的MapReduce化等问题,并给出了有效的解决方案。最后,本文在上文的理论支持和对云应用开发的相关技术充分调研的基础之上,搭建了Hadoop分布式集群环境,并且实现了基于Hadoop云平台的分布式视频关键帧提取系统.实验结果表明,本文算法能够很大程度的提高关键帧的提取速度,与单机提取模式相比,更加适合处理视频大数据。
[Abstract]:With the video surveillance technology has been widely used in various industries, to monitor the amount of video data is exponential growth, it has already tired of the previous time according to the information manually drag the progress bar to browse, query not only consumes a lot of time and energy, but also easy to miss the target information. Key technology of key frame extraction the situation has been greatly improved, the key frames can be not affected by time, influence of synchronization of audio and video, and can provide a variety of ways for browsing and navigation. But in the actual application process of extraction of key frames, the biggest difficulty is the key frame extraction speed is too slow: on the one hand is the high complexity algorithm; on the other hand is the current use of method is stand-alone mode. Based on the full investigation of key frame extraction algorithm, put forward a new turn Key frame extraction method, and the algorithm is transplanted to the Hadoop cloud platform, solve the frame integrity of the core issues, success will be transformed into a distributed algorithm extraction method, greatly improving the extraction rate of key frame. Firstly, this paper points out the importance of practical engineering technology of key frame of video retrieval applications in recent years, and introduces the development like a raging fire of cloud computing technology and its application in the field of multimedia, the core technology and theory of Hadoop are full of learning and research. Secondly, this paper analyzes the current problems of key frame extraction method, video key frame extraction method is proposed and a new adaptive. This method can adaptively determine the number of key frames, better treatment effect and small amount of calculation of the content gradient of the video. Thirdly, this paper analyzes on the Hadoop cloud platform to realize the video off Distributed key frame extraction problem faced by cloud applications, such as frame integrity, MapReduce issues such as processing logic, and gives effective solutions. Finally, based on the above theoretical support and related technology of cloud application development of full investigation, set up a Hadoop distributed cluster environment, and the implementation of distributed video key frame extraction system based on Hadoop cloud platform. The experimental results show that this algorithm can improve the speed of extracting key frames greatly, compared with the single extraction mode, more suitable for processing video data.

【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN948.6;TP393.09

【相似文献】