Short Video Retrieval System Based On Mobile Platform

There are two primary methods for video retrieval,one is text based video retrieval(TBVR)and the other is content based video retrieval(CBVR).Because that the content based video retrieval system needs less labor cost and reach closer to users’ demands,CBVR systems become the mainstream of research topics about the video retrieval system.Recent years,the development of mobile makes the users of mobile video increase extremely fast.In all video systems of mobile platform,the original short video become the major force.But there are limitations of the traditional TBVR for the flexibly and irregularity of original short video.So we propose a content based video retrieval system based on the short videos on the mobile platform.Our system needs the user input the image to be searched.The server returns the result and the user can use the APP to broadcast the videos.Our system has fast retrieval speed.Through the experiment,we prove that the result is accurate and the data transmission between server and client is stable.We also reach the good user experience.We build the whole system,including the architectural design,the algorithm of research,the coding of server and client platform.Besides that,we also build a dataset to judge our system.In this paper,we described the system architecture and introduce the detailed design of server,client and network.The steps of the retrieval can be divided to four steps: coalesce the optical flow motion analysis and the reference frame analysis for key frame extraction,use the color and edge directivity description(CEDD)to extract the feature,build the Lucene index of the extracted feature and search the videos.We also analyze the dataset of experiment videos to build the description of our dataset.