一种基于关键帧的分布式视频分析解耦机制  

Distributed Video Analysis Decoupling Mechanism Based on Keyframe

作　　者：蔡晓东[1] 华娜[1] 吴迪[1] 陈文竹[1] 

机构地区：[1]桂林电子科技大学信息与通信工程学院,广西桂林541004

出　　处：《电视技术》2015年第14期1-4,20,共5页Video Engineering

基　　金：国家科技支撑计划项目(2012BAH20B01;2014BAK11B02);广西自然科学基金项目(2012GXNSFAA053232;2013GXNSFAA019326);广西高校科学技术研究项目(2013YB092)

摘　　要：近年来云计算逐渐应用于海量视频数据分析,但视频的处理单位是视频帧,而视频帧的大小是不固定的,所以不能直接用云平台处理视频,否则会出现帧不完整、找不到视频头文件、视频解码不成功等情况。通常的解决办法是对视频数据进行预处理,将视频分成若干独立可解码的视频块提交到文件系统中进行处理,但是当数据量很大时,预处理的工作量也很大。因此提出了一种基于关键帧的分布式视频分析解耦机制,将文件系统中的视频块按关键帧位置对视频数据进行巧妙分片。实验表明,提出的分布式视频分析解耦机制可有效对视频进行并行处理,保证了帧的完整性,且与云平台默认的分块方式相比计算负载更加均衡,提高了整体处理效率。In recent years, cloud computing is used in the analysis of massive video data, while the unit of video processing is a frame, and the frame size is not fixed, so it call not directly proeess video in cloud platform, otherwise the following events occur, such as no complete fi＇ame, no header file, unsuccessful video decoding. Usually the solution is video preprocessing, videos are divided into several independent video blocks and video blocks are submitted to the file system, while the pre-processing workload will be greater when the amount of data is large. So it proposes a distributed video analysis deeoupling mechanism based on keyframe, to fragment the video blocks in file system by keyframe positions. Experiments show that the proposed mechanism fo, distributed video decoupling can be effectively process video in parallel, and ensure the integrity of the frame, computational load is more balanced, the way improves tile overall process efficiency.

关 键 词：云平台 关键帧 分布式 解耦 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]