Design And Implementation Of Video Classification Based-on Distributed Platform

The rapid of development of Internet and intelligent mobile devices has already led to a large amount of data generated over the past two decades,and multimedia data such as videos accounts for a big percentage of this deluge of data. According to Youtube video platform 2016 statistics show that Youtube per minute to receive the user to upload, has reached a 300 hours of playing time. In the 2015-2020 consumer Internet traffic report,Cisco point that the Internet video streaming and downloads are beginning to take a larger share of bandwidth and will grow to more than 80 percent of all consumer Internet traffic by 2020. In order to organize these videos and discover useful information, it is worth understanding human activities automatically from videos. To meet the escalating computational demands,the development of distributed applications deployed on computer clusters has become much productivity since a wide range of programming paradigms, middle-wares and framework are developed.In this paper, Spark, the rising big data processing tool with in-memory computing ability, is explored to address the task of large-scale human action recognition. To achieve this, several key techniques for human action recognition, such as K-means cluster, VLAD encode algorithm,Fisher Vector encode algorithm etc., are realized with parallel distributed computing power on Spark. The theory and implementation details for these distributed applications are presented in this work.